Record for a UF thesis. Title & abstract won't display until thesis is accessible after 2013-08-31.

MISSING IMAGE

Material Information

Title:
Record for a UF thesis. Title & abstract won't display until thesis is accessible after 2013-08-31.
Physical Description:
Book
Language:
english
Creator:
Walkup,Brian Ross
Publisher:
University of Florida
Place of Publication:
Gainesville, Fla.
Publication Date:

Thesis/Dissertation Information

Degree:
Doctorate ( Ph.D.)
Degree Grantor:
University of Florida
Degree Disciplines:
Business Administration, Finance, Insurance and Real Estate
Committee Chair:
Ryngaert, Michael D
Committee Co-Chair:
Nimalendran, Mahendrara
Committee Members:
Houston, Joel F
Hamersma, Sarah E

Subjects

Subjects / Keywords:
Finance, Insurance and Real Estate -- Dissertations, Academic -- UF
Genre:
Business Administration thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Statement of Responsibility:
by Brian Ross Walkup.
Thesis:
Thesis (Ph.D.)--University of Florida, 2011.
Local:
Adviser: Ryngaert, Michael D.
Local:
Co-adviser: Nimalendran, Mahendrara.
Electronic Access:
INACCESSIBLE UNTIL 2013-08-31

Record Information

Source Institution:
UFRGP
Rights Management:
Applicable rights reserved.
Classification:
lcc - LD1780 2011
System ID:
UFE0043211:00001


This item is only available as the following downloads:


Full Text

PAGE 1

1 TWO ESSAYS IN FINANCE By BRIAN R. WALKUP A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2011

PAGE 2

2 2011 Brian R Walkup

PAGE 3

3 To my loving wife Tracy and to my wonderful parents Your love and support has not gone unnoticed

PAGE 4

4 ACKNOWLEDGMENTS I would like to thank everyone that has helped support me throughout this long journey. I thank my parents for always being behind me 100% throughout my entire education and my life I thank my beautiful wife, Tracy, for being there to get me through the tough times and for never doubting me or allowing me to doubt myself I thank my Committee, Michael Ryngaert (Chair), Mahendrarajah Nimalendran (Coc hair), Joel Houston and Sarah H a mersma for their significant contributions as well as countles s hours of support, emails and conversations regarding my dissertation. I thank my fellow f inance Ph. D students for helping make the process enjoyable.

PAGE 5

5 TABLE OF CONTENTS page ACKNOWLEDGMENTS .................................................................................................. 4 LIST OF TABLES ............................................................................................................ 7 LIST OF FIGURES .......................................................................................................... 9 ABSTRACT ................................................................................................................... 10 CHAPTER 1 INTRODUCTION .................................................................................................... 12 Overview of Chapter 2 ............................................................................................ 12 Overview o f Chapter 3 ............................................................................................ 14 2 THE EFFECTS OF UNCERTAINTY AND TAXES ON CORPORATE PAYOUT POLICY ................................................................................................................... 16 Data ........................................................................................................................ 23 Sample Creation ............................................................................................... 23 Dependent Variables ........................................................................................ 24 Explanatory Variables ...................................................................................... 26 Summary Statistics ........................................................................................... 35 Results .................................................................................................................... 36 Payout Policy Dividends ................................................................................ 36 Payout Policy Repurchases ........................................................................... 44 GARCH Robustness Check .................................................................................... 48 Conclusion .............................................................................................................. 51 3 PRICE DISCOVERY AND RECENT TRENDS IN EXTENDEDHOURS TRADING ................................................................................................................ 78 Overview of the Existing ExtendedHours Trading Literature ................................. 81 Data ........................................................................................................................ 85 Trends in the ExtendedHours Trading Environment .............................................. 86 Price Discovery in ExtendedHours Trading ........................................................... 89 Conclusion .............................................................................................................. 9 3 4 CONCLUSION AND FUTURE WORK .................................................................. 109 APPENDIX: GARCH (1,1) ESTIMATE OF THE VIX ................................................... 112

PAGE 6

6 LIST OF REFERENCES ............................................................................................. 131 BIOGRAPHICAL SKETCH .......................................................................................... 134

PAGE 7

7 LIST OF TABLES Table page 2 1 Variable definitions. ............................................................................................ 54 2 2 Summary statistics for the whole sample ........................................................... 56 2 3 Summary statistics for dividend nonpayers verse dividend payers ................... 57 2 4 Summary statistics for nonrepurchasers verse repurchasers ............................ 58 2 5 Logit regression for the choi ce to pay or not pay a dividend ............................... 59 2 6 Linear combination for the choice to pay or not pay a dividend .......................... 61 2 7 Multinomial logit regressi on for the choice to maintain or change dividend for prior dividend payers .......................................................................................... 62 2 8 Linear combination for the choice to maintain or change dividend for prior dividend payers .................................................................................................. 64 2 9 Logit regression for the choice to maintain or pay a dividend for prior dividend n onpayers .......................................................................................................... 65 2 10 Linear combination for the choice to maintain or pay a dividend for prior dividend nonpayers ........................................................................................... 67 2 11 Logit regression for the choice to repurchase shares or not repurchase shares ................................................................................................................. 68 2 12 Linear combination for the choice to repurchase shares or not repurchase shares ................................................................................................................. 70 2 13 Multinomial logit regression for the choice to maintain or change repurchase level for prior repurchasers ................................................................................. 71 2 14 Linear combination for the choice to maintain or change repurchase level for prior repurchasers .............................................................................................. 73 2 15 Logit regression for the choice to initiate a repurchase for prior nonrepurchasers ............................................................................................... 74 2 16 Linear combination for the choice to maintain or increase repurchase level for prior nonrepurchasers ....................................................................................... 76 3 1 Average monthly extendedhours turnover by year ............................................ 96

PAGE 8

8 3 2 Number of extendedhours trades per year and average extendedhours trade size ............................................................................................................ 97 3 3 Average daily number of extendedhours trades by 30minute interval .............. 98 3 4 Percentage of extendedhours trades occurring by venue ................................. 99 3 5 Weighted Price Contribution by year ................................................................ 100 3 6 Absolute Price Discovery by year ..................................................................... 101 A 1 Logit regression for the choice to pay or not pay a dividend ............................. 113 A 2 Linear combination for the choice to pay or not pay a dividend ........................ 115 A 3 Multinomial logit regr ession for the choice to maintain or change dividend for prior dividend payers ........................................................................................ 116 A 4 Linear combination for the choice to maintain or change dividend for prior dividend payers ................................................................................................ 118 A 5 Logit regression for the choice to maintain or pay a dividend for prior dividend nonpayers ........................................................................................................ 119 A 6 Linear combination for the choice to maintain or pay a dividend for prior dividend nonpayers ......................................................................................... 121 A 7 Logit regression for the choice to repurchase shares or not repurchase shares ............................................................................................................... 122 A 8 Linear combination for the choice to repurchase shares or not repurchase shares ............................................................................................................... 124 A 9 Multinomial logit regression for the choice to maintain or change repurchase level for prior repurchasers ............................................................................... 125 A 10 Linear combination for the choice to maintain or change repurchase level for prior repurchasers ............................................................................................ 127 A 11 Logit regression for the choice to maintain or increase repurchase level for prior nonrepurchasers ..................................................................................... 128 A 12 Linear combination for the choice to maintain or increase repurchase level for prior nonrepurchasers ..................................................................................... 130

PAGE 9

9 LIST OF FIGURES Figure page 2 1 Percentage of firms decreasing/eliminating dividends vs percentage of firms increasing dividends by year .............................................................................. 77 3 1 Turnover by month for the S&P 500, S&P 400 and S&P 600 ........................... 102 3 2 Number of extendedhours trades by month .................................................... 103 3 3 Size of extendedhours trades by month .......................................................... 104 3 4 Number of extendedhours trades by 30minute interval .................................. 105 3 5 Percentage of extendedhours t rades by v enue. .............................................. 106 3 6 Weighted Price Contribution by year ................................................................ 107 3 7 Absolute Price Discovery by year ..................................................................... 108

PAGE 10

10 Abstract of Dissertat ion Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy TWO ESSAYS IN FINANCE By Brian R Walkup August 2011 Chair: Michael Ryngaert Cochair: Mahendrarajah Nimalendran Major: Business Administration In this study I look at two very distinct topics in finance. The first part of the study examines the impact of market wide uncertainty on corporate payout policy. While the prior literature has mostly focused on internal, firm level determinants of payout policy, I show that market wide economic uncertainty (as proxied for by the VIX) also plays a significant role in the payout policy decision. By utilizing interaction variables for each of the volatility measures I am able to demonstrate that the impact of both internal and external volatility measures depends upon the firms level of cash flow. Firms with relatively low cash flow are significantly more likely to cut dividends when market level volatility is higher than are firms with relatively high cash flow. With regards to the decision to repurchase stocks I show that high cash flow firms become opportunistic during highly volatile markets and are more likely to initiate a repurchase. Companies with high firm return volatility tend to cut back on repurchases. I also demonstrate th at tax changes, such as dividend tax rate changes and repatriation tax cuts, have an impact on payout policy. In the second part of the study I examine the changes that have occurred in extendedhours trading since noninstitutional traders were first given access in 1999.

PAGE 11

11 Using a large sample of extendedhours trades from 1999 through 2009, I find that the trends identified in regular trading hours do not necessarily transfer to the extendedhours trading sessions. I also utilize both the Weighted Price Contribution measure as well as a newly created measure, Absolute Price Discovery, to examine changes in the portion of daily price discovery that occurs outside of regular trading hours over time. I show that extendedhours trading has become significantly more important to the price discovery process over my sample period. For S&P 1500 Composite Index firms the percentage of price discovery occurring during extendedhours trading has risen from 5.71% in 1999 to 25.28% in 2009. For the larger stocks that comprise the S&P 500 Index this growth is even more dramatic, from 6.05% in 1999 to 41.09% in 2009.

PAGE 12

12 CHAPTER 1 INTRODUCTION In the two chapters that constitute this study I examine two very distinct topics in the field of finance: payout policy and extendedhours trading The first part of the study, Chapter 2, investigates the impact market level uncertainty has on a firms payout policy decision. While m uch of the prior literature on the determinants of payout policy focuses on internal, firm level factors, I show that external factors can al so have descriptive power on the payout choice of firms even after controlling for firm level characteristics. Using the Chicago Board Options Exchange Volatility Index (VIX) as a proxy for market level uncertainty, I demonstrate that firms with low level s of cash flow become more conservative with their payout policy as indicated by an increasing propensity to halt div idend increases or even decrease or eliminate their dividend altogether. With regards to repurchases I show that firms with high levels of cash flow become opportunistic by increasing the likelihood of initiating a share repurchase. In the second part of the study, Chapter 3, I shift the focus to extendedhours trading. I examine the trends that have occurred since the introduction of noninstitutional investors to the after hours trading environment in 1999. As the United States financial market becomes increasingly intertwined with the global market and as information flows across the 24hour day become more seamless, I demonstrate that the importance of extendedhours trading on the price discovery process has increased significantly. Overview of Chapter 2 The prior literature on the determinants of payout policy has focused almost entirely on internal firm characteristics. In Chapter 2 I demonstrate that external,

PAGE 13

13 market level factors can also impact the decision of a firm with regards to dividend payment or share repurchases. Using the VIX as a proxy for market level uncertainty I find the differing affect on the payout policy decision (both dividends and repurchases) depend e nt on the firms relative cashflow levels. Even after controlling for a wide range of previously utilized firm level determinants of payout policy, I show that the VIX has a significant impact. Lintners (1956) dividend stickiness theory argues that firms are reluctant to reduce dividend payout levels due to the perceived negative signal and resulting stock price decrease that is associated with a dividend decrease. I show that firms with relatively low levels of cash flow choose to become conservative in their approach to cash holdings when market level uncertainty is high. They become less likely to increase their dividend over preestablished levels and have an increased probability of decreasing or even eliminating their dividend. Firms with relatively high levels of cash flow are better suited to withstand these times of market volatility While they do still become somewhat more conservative, as evidenced by a decreased likelihood of a dividend increase, they do not become significantly more likely to decrease or eliminate their dividend when market wide volatility is high. G iven that changes in the level of stock repurchase are not viewed by investors or firms as the same type of long term commitment as dividend changes, I show a very different impact of market level volatility on the repurchase decision. F irms with high levels of cash flow are actually more likely to initiate a repurchase during high VIX periods. This result likely represents these firms utiliz ing periods of high volatility to opportunistically repurchase shares at a low level The firm may be able to identify

PAGE 14

14 periods of underpricing for their shares due to the high volatility or may time repurchases for periods where investors require high market risk premiums therefore resulting in temporarily low stock prices. Fir ms with r elatively low levels of cash flow on the other hand, are not significantly affected by the uncertainty in terms of the repurchase decision. This may reflect a lack of cash flow to repurchase opportunistically. Overview of Chapter 3 Extendedhours trading refers to trading that occurs outside of the normal 9:30 am to 4:00 pm trading day. Noninstitutional traders were first given access to trade during the extendedhours in 1999. In Chapter 3 I examine the general trends in extendedhours trading since 1999 with specific attention being paid to the price discovery process. I show that the trend towards a larger volume of smaller trades documented by Chordia, Roll and Subrahmanyam (201 1 ) during regular trading hours does not necessarily transfer to extended hours trading. While the volume of trading occurring outside of trading hours has grown over time, the size of the trades has not shown a significant drop over an extended period as has been seen in regular trading hours. Using both the previously es tablished Weighted Price Contribution (WPC) measure of Barclay and Hendershott (2003) and my own newly created Absolute Price Discovery (APD) measure, I demonstrate that a significant portion of the 24hour price discovery process has shifted from regular trading hours for U.S. stock exchanges and is now occurring before 9:30 am and after 4:00 pm. The percentage of price discovery occurring outside of trading hours has steadily increased from approximately 8.49% in 1999 to approximately 32.35% in 2009 using the WPC for the S&P 1500 Composite

PAGE 15

15 Index. Utilizing the APD, a newly created variable that adjusts for a potential upward bias in the WPC, the percentage of price discovery taking place in extendedhours trading still increased from 5.71% in 1999 to 25. 28% in 2009. This increase in price discovery over time is even more dramatic for larger, more heavily trade stocks. Breaking the sample down to only the S&P 500 largecap stocks reveals an increase in WPC from 10.11% in 1999 to 57.52% in 2009. The incr ease using the APD measure is from 6.05% to 41.09%. This shift in the price discovery process represents changes in the United States financial market as it becomes more dependent on changes in the global economy and as information flow through t he day becomes easier and more continuous. Though extendedhours trading has been a mostly ignored area in the academic finance literature, I believe that this study demonstrates its growing importance. Given the significant growth in price discovery out side of trading hours it is clearly not of marginal importance. In future extensions I plan to continue to fill the gap in the literature regarding extendedhours trading.

PAGE 16

16 CHAPTER 2 T HE EFFECTS OF UNCERT AINTY AND TAXES ON C ORPORATE PAYOUT POLI CY Should firms pay out a portion of their earnings in the form of dividends? If so, what percentage of earnings should be paid out and how should this level be determined? While a significant portion of the financial payout policy literature has debated the divi dend puzzle, as articulated by Black (1976) which questions why firms pay dividends when they appear to be tax disadvantaged, another large portion has simply accepted the fact that a significant number of firms do pay dividends and have attempted to show the determinants that affect the dividend decision for firms.1 In general these studies have utilized firm level attributes to demonstrate the types of firms that pay dividends and the characteristics that may lead to changes in dividend policy. One such study is Fama and French (2001) who argue that dividends are disappearing over the period 1963 to 1998. The authors show that the percentage of firms paying dividends decreased from a high of nearly two out of every three firms (66.5%) in 1978 to approxi mately one in five firms (20.8%) in 1999.2 Fama and French (2001) try to explain the disappearance of dividends due to the changing composition of publicly traded firm types In general large firms with more steady earnings are significantly more likely to pay dividends than their smaller, more volatile counterparts. As the composition of publicly traded firms shifted towards more small, growthoriented 1 Fischer Black coined the phrase dividend puzzle in his 1976 paper entitled simply The Dividend Puzzle. In this paper, Black considered arguments described in prior literature, specifically Miller and Modigliani (1961), which set forth a model in whic h dividends were irrelevant given an efficient market with no transaction costs, no bankruptcy costs and no asymmetric information. 2 DeAngelo, DeAngelo and Skinner (2004) show that, while the number of dividend payers did decrease, the aggregate level of real dividends paid by industrial firms actually increased over this time. This is due to the high percentage of total dividends being paid out by the largest dividendpaying firms and the relative small amount of dividends that were being paid by the fi rms which reduced or eliminated dividends.

PAGE 17

17 firms near the end of the Twentieth C entury it was logical that there would be a decrease in the proportion of firms paying steady dividends. However, Fama and French (2001), even after accounting for these firm level attributes, still find that the likelihood of a firm paying out dividends decreased during this time period. While Fama and French (2001) focus on firm level attributes, they do not account for factors external to the firm when explaining dividend choice. Baker and Wurgler (2004) explore this possibility by looking at the potential impact that external factors could have on payout policy. According to their Catering Theory, the premium that investors assign to dividendpaying firms changes over time. These shifts in dividend premiums can be affected by a variety of factors including changes in tax rates and changes in investor sentiment levels. The Catering Theory would imply that, as the value investors place on the payment of dividends shifts over time, firms are willing to shift their payout policy to cater to investors preferences.3 Though Baker and Wurgler (2004) find that their e stimated dividend premium does influence firms propensity to initiate dividends during their sample period, they do not find that the dividend premium has a significant impact on the propensity to continue dividends for firms which have already paid a div idend in the prior fiscal year. Chay and Suh (2009) examine the impact of firm level uncertainty as a determinant of payout policy. Prior survey evidence (such as Brav et al ( 2005) ) argues that a firms level of cashflow uncertainty plays a large part in the decision of whether or not to change the dividend level, though the exact operational definition of this 3 The Catering Theory can be viewed as an extension of the Miller Modigliani (1961) notion of dividend clienteles in a world with market frictions that can slow the adjustment of the supply of dividends to the dem and for them.

PAGE 18

18 uncertainty is not surveyed. Chay and Suh (2009) confirm this empirically. Using monthly stock return volatility as a proxy, they show that a fi rms stock return volatility is a strong determinant of its payout policy. This finding ties in well with the dividend stickiness theory of Lintner (1956) which argues that firms are hesitant to make major changes in dividend policy to avoid the stock pri ce penalty that is associated with decreasing or eliminating their dividend. Similar to Chay and Suh (2009), Hoberg and Prabhala (2009) look at how uncertainty affects the payout decision. However, Hoberg and Prabhala (2009) use idiosyncratic risk and fir m level systematic risk as their measures of uncertainty. This allows them to not only capture the firms diversifiable risk, but also the nondiversifiable risk. After including their risk factors, Hoberg and Prabhala (2009) show that the dividend premi um from Baker and Wurgler (2004) is no longer significant, which leads them to question the importance of the Catering Theory. In this paper I build on the framework set up by Fama and French (2001), Baker and Wurgler (2004), Chay and Suh (2009) and Hoberg and Prabhala (2009). I show that another significant determinant of a firms payout policy choice is market wide uncertainty that is external to the firm as defined by the popular risk measure, the Chicago Board Options Exchange Volatility Index (VIX); t he implied volatility from options on the S&P 500. While Baker and Wurgler (2004) begin to touch on the fact that external, nonfirm specific factors may have an impact on the payout policy of firms, they do not fully explore what these factors may be, such as the VIX. On the other hand, Chay and Suh (2009) and Hoberg and Prabhala (2009) show that uncertainty

PAGE 19

19 plays a role in the payout decision, but only examine individual firm volatility.4 I bridge this gap and show that external market level uncertainty is actually a significant factor when a firm is making its payout policy decisions. When market conditions become less stable all firms are likely to become less confident in their ability to maintain an appropriate level of free cash to sustain current dividend levels. Therefore dividend paying firms become more likely to decrease/eliminate dividends and conserve cash. At the same time non dividend paying firms are less likely to initiate dividends during volatile market wide economic conditions. On the other hand, when the market wide volatility is lower, even firms with relatively high internal cash flow uncertainty may feel more comfortable initiating a low level of dividends. Figure 2 1 compares the percentage of firms decreasing/eliminating div idend levels relative to the percentage of firms increasing dividend levels over time. It becomes quite apparent that firms have been more likely to decrease/eliminate during highly uncertain market conditions (near the bursting of the technology bubble i n the early 2000s and again during the subpr ime mortgage crisis around 2007 to 2009). Whether this is driven by declines in profitability or reactions to extreme market uncertainty is an empirical question. The recent market instability associated with the subprime mortgage crisis is a very good example of the effect market conditions can have on payout policy. As credit becomes less available to firms and earnings begin to fall dividends become less attractive to firms, particularly if there is heightened uncertainty about the future. This 4 Hoberg and Prabhala (2009) do examine firm idiosyncratic risk and systematic risk from a market model. Their systematic risk component, however, has a firm specific component (the beta of the stock), is not forwardlooking like the VIX and arguably the VIX may capture other elements of the market fear of uncertainty. In fact, the VIX is often viewed as a measure of market fear. See, for example, Whaley (2000) and Arak and Mijid (2006) for further details on the VIX and its commonly us ed nickname the fear gauge or fear index.

PAGE 20

20 can be seen clearly in the press releases issued at the time of dividend announcements over the past few years. Below is a small sampling of quotes from press releases d uring the subprime mortgage crisis: International Paper Company: Decreased dividend from 25 cents per share to 2.5 cents per share o Direct quote from Chairman and CEO John Faraci: While our cash balances and cash flows remain solid, we believe it is prudent to manage cash conservatively in this uncertain economic environment. (Reuters) Entercom Communications Corp: Eliminated prior dividend of 10 cents per share after having already decreased the dividend from 38 cents per share in a prior quarter o has su spended the Companys dividend in light of the difficult business environment and the uncertain outlook for the U.S. economy. (Business Wire) Dover Motorsports: Decreased dividend from 1.5 cents per share to 1 cent per share and later eliminated their div idend in a subsequent quarter o The company believes that adjusting the dividend is prudent given the current economic environment and will afford it greater financial flexibility moving forward. (Business Wire) Kenneth Cole: Eliminated prior dividend of 9 cents per share after having already decreased the dividend from 18 cents per share in a prior quarter o The company said it is suspending its 9 cent dividend to preserve and manage liquidity in a highly uncertain environment. (Associated Press) JP Morg an Chase & Co: Decreased dividend from 38 cents per share to 5 cents per share o Extraordinary times call for extraordinary measures. (Dow Jones News Service) To account for the impact of market level economic conditions I utilize the Chicago Board Options Exchange Volatility Index (VIX) as a proxy for market level volatility and uncertainty. The VIX measures implied volatility on the S&P 500 index options and is

PAGE 21

21 often referred to as the fear index for the general market.5 I show that the VIX adds power to tests of the probability of a firm changing their dividend policy even in the presence of other uncertainty measures such as firm stock return volatility. In addition to the VIX I also introduce a measure of analyst dispersion as another volatility variable for the payout choice. Dispersion of analyst opinion captures the uncertainty of the market regarding the firms future earnings taking into account both firm level and market level factors. It is a strong forwardlooking measure and thus is potentially important to include in tests of payout choices. I show that firms are significantly more (less) likely to decrease or eliminate (increase or initiate) their dividends when analyst dispersion is high (low). Another contribution of this study is t hat I show how firms in different relative cash flow positions react differently to both firm level stock volatility and market wide volatility. In general, firms with low relative cash flow become more sensitive to external uncertainty while firms with high relative cash flow are not as severely affected. I also utilize two measures of tax rate changes as control variables from the prior payout literature. I include a variable that measures the change in the dividend tax rate. This variable allows my m odel to capture the effect of dividend tax rate changes on dividend policy. The prior literature, such as Chetty and Saez (2005), shows that large dividend tax changes can have a significant impact on payout policy.6 The second tax rate variable captures the effect of repatriation tax cuts for firms that have positive foreign 5 See Whaley (2000) and Arak and Mijid (2006) for further details about the VIX and why it is commonly referred to as the fear gauge or fear index. 6 Brav, et al (2008) survey 328 financial executives and demonstrate that, while not the most important factor in the payout policy decision, the 2003 dividend tax cut did play a significant role in the decision by many firms to initiate or increase dividends.

PAGE 22

22 income. Blouin and Krull (2009) demonstrate that repatriating firms increase repurchases significantly more than nonrepatriating firms in 2005 in response to the American Job Creation Act. Therefore it seems sensible to believe that it may have some impact on dividend policy as well. When combined with control variables from prior research (size, market to book, profitability and growth rate of assets from Fama and French (2001) and cash flow volatility from Chay and Suh (2009)), the VIX, analyst dispersion and tax rate measures help create a much more powerful test of dividend policy choice. Since Skinner (2008) shows that firms are increasingly using repurchases both as a comple ment and a substitute to dividends, each test conducted for dividends is replicated for the choice to repurchase and the decision to change repurchasing levels. Again, it is worthwhile to also look at the impact that internal and external volatility, thei r interaction with relative cash flow levels, and the taxation variables have on the repurchase choice. In general, firms with high internal volatility show mixed results, but appear to be more reluctant to conduct or initiate share repurchases, particularly at higher cash flow levels. It also appears that firms with high cash flow levels are more likely to initiate or conduct repurchases when market uncertainty (VIX) is high. This may be evidence of opportunistic behavior by high cash flow firms. Addit ionally, evidence tends to show that firms more freely switch their repurchasing level to match their internal earnings, cash flow, volatility, etc. This is consistent with prior literature and beliefs that repurchases are much less sticky than dividends and repurchase levels are more freely switched than are dividend levels.7 Furthermore, firms that are 7 Brav, et al (2005) show in survey evidence that f irms are more likely to use new free cash flow for repurchases than dividends due to the sticky nature of dividends discussed in Lintner (1956)

PAGE 23

23 likely to bring back cash due to repatriation taxes are more likely to increase repurchases, consistent with them potentially viewing such an event as a onetime change that is more conducive to doing a repurchase rather than committing to a higher dividend level. Given the fact that my original sample only looks at 1990 to 2009 due to the VIX measure only being available post 1990, there is concern that my results may be overly influenced by the high VIX period which occurred during the recent financial crisis. To ensure that the results are robust to both a longer sample period and to removing the recent crisis, I construct a generalized autoregressive conditional heteroskedasticity (GARCH) model to estimate market volatiltiy back to 1962. This GARCH estimate has a correlation of 0 .913 with the VIX through the overlapping period of 1990 to 2009. I then replace the VIX measure with the GARCH measure and rerun all tests in this study. I find that the main results hold in both the full sample (1962 to 2009) and the nonoverlapping sample (1962 to 1989). This demonstrates that the results ar e not caused solely by the high VIX period of the financial cris is. Data Sample Creation The data set used in this study is for the period 1990 to 2009. The initial sample is calculated in a manner very similar to Fama and French (2001). I started with the universe of firms covered by Compustat and eliminated all uti lities (firms with Standard Industrial Classification (SIC) Codes between 4900 and 4949) and financial firms (firms with SIC Codes between 6000 and 6999). Firms were required to have book equity (defined as stockholders equity (Compustat Item #216) minus preferred stock (Compustat Item #10) plus deferred taxes and investment tax credits (Compustat Item

PAGE 24

24 #35) plus post retirement assets (Compustat Item #15)) greater than $250,000 and total assets (Compustat Item #6) greater than $500,000 to be included in t he sample. All firms in the sample were also required to have nonmissing values for common shares outstanding (Compustat Item #25), total assets (Compustat Item #6), price close fiscal (Compustat Item #24), income before extraordinary items (Compustat I tem #18), dividends per share ex date fiscal (Compustat Item #26), interest and related expense (Compustat Item #15), preferred dividends (Compustat Item #19), and earnings before interest taxes, depreciation and amortization. To be included in the s ample firms must also have either (a) stockholders equity (Compustat Item #216), (b) common equity (Compustat Item #60) and preferred stock (Compustat Item #130), or (c) total liabilities (Compustat Item #118) as well as either (a)preferred stock / liquidating value (Compustat Item #10), (b) preferred stock / redemption value (Compustat Item #56), or (c) preferred stock (Compustat Item #130). I obtained Standard Industrial Classification (SIC) Codes from CRSP as well as share code data. To remain in the sample firms must have share codes of either 10 or 11 to ensure that they are publicly traded. Dependent Variables In this study I look at the payout choice in three different specifications. The first specification is simply the choice to either pay or not to pay a dividend. The second specification is the choice to increase, decrease or maintain current dividend levels, given that the firm paid a dividend in the prior fiscal year. The final specification I utilize is the choice to maintain no dividend or initiate a dividend given that the firm did not pay a dividend in the prior fiscal year. This same set of specifications is also utilized replacing dividends with repurchases.

PAGE 25

25 For the first specification style, a firm is considered to be a payer if the ex date dividend per share (Compustat Item #26) is positive. Similarly, when utilizing the first specification for repurchases, a firm is considered to be a repurchaser if total repurchases are positive. Repurchases are defined similarly to Grullon and Michaely (2002) as purchase of common and preferred stock (Compustat Item #115) minus any reduction in the redemption value of preferred stock (Compustat Item #56).8 Banyi, Dyl and Kahle (2008) show that this measure is the most accurate of the commonly utilized repurchase definitions. For the second specification I look at the choice set for firms that have paid a dividend in the prior fiscal year. I utilize a one fiscal year lagged value of the payer term used in the first specification to identify fir ms that paid a dividend in the prior fiscal year. A firm is considered to have decreased (increased) their dividend level if the dividend per share for the current fiscal year has decreased (increased) more than 5% relative to the prior fiscal year. If t he dividend per share stays within 5% of the initial dividend per share level of the prior fiscal year then the firm is considered to have maintained their prior dividend policy. The same definitions are utilized for repurchases, except replacing dividends per share with the repurchase definition (as stated in prior paragraph) and replacing 5% with 20%. The wider range for the definitions of increasing/decreasing on repurchasing is utilized because firms repurchase policies are generally less stable than dividend policies. By utilizing a wider range I am better able to identify the characteristics that result in a significant change in policy. 8 Reductions in the redemption value are required to be nonnegative. Therefore, reduction in the redemption value of preferred stock (defined as Compustat Item #56 minus the one fiscal year lagged value of Compustat Item #56) is truncated at 0.

PAGE 26

26 The final specification considers firms which did not pay a dividend in the prior fiscal year. If the dividend per share for the current fiscal year is positive then the firm is considered to have initiated a dividend. If not, the firm is considered to have maintained the prior payout policy to not pay a dividend. Each of the three specifications is also conducted for the repurchase choice. In these specifications dividend per share is replaced with net repurchases. Explanatory Variables The explanatory variables utilized in my regressions can be grouped into four specific categories: (1) Fama and French (2001) variables, (2) internal and external volatility measures, (3) tax measures, and (4) additional firm level controls. Fama and French (2001) variables. The subset of Fama and French (2001) variables consists of the four variables primarily utilized in the Fama and French (2001) study. These four variables are chosen as a base for the regressions as they are commonly accepted explanatory variables for the payout choice. The first variable in this category is NYP which is defined as the percentage of firms on the New York Stock Exchange with the same or lower market capitalization relative to the firms market capitalization for the current fiscal year. This is calculated in 5% intervals.9 NYP captures the size effect as larger, more established firms are generally more likely to pay dividends. The next variable in the Fama and French (2001) subset of explanatory variables is MtoB which represents the firms market to book ratio for the current fiscal year. The 9 For example, if firm XYZ has a market capitalization in fiscal year t which is equal to or less than the cutoff for the bottom 5% of the NYSE market capitalizations than the corresponding value of NYP for firm XYZ in fiscal year t is 0.05.

PAGE 27

27 market to book ratio is defined as (total assets (Compustat Item #6) minus book value plus market equity) divided by total assets. Book value is defined as stockholders equity (Compustat Item #216) minus preferred stock (Compustat It em #10) plus deferred taxes and investment tax credits (Compustat Item #35) plus post retirement assets (Compustat Item #15). Market equity is defined as the stocks closing price (Compustat Item #24) multiplied by the number of common shares outstanding (Compustat Item #25). Fama and French (2001) utilize MtoB as one measure to identify investment opportunities. The third Fama and French (2001) variable is dA/A which represents the change in assets (lag total assets total assets) divided by lag total assets. As with MtoB dA/A is used to identify investment opportunities. In both cases, we should expect higher investment opportunities to correlate with decreased probability of dividend payment. The final variable from Fama and French (2001) is E/A which is defined as (income before extraordinary items (Compustat Item #18) plus interest and related expenses (Compustat Item #15) plus deferred incomes taxes (C ompustat Item #50)) divided by total assets. A higher level of earnings allows for more opportunity to pay dividends to investors. Internal and external volatility measures. In this study I aim to empirically show that payout policy is affected by not only internal, firm level volatility (as shown in Chay and Suh (2009)), but also by external, market level volatility. For my measure of firm level volatility I utilize ReturnVolatility defined as the annualized monthly standard deviation of stock returns f or the 24 month period including the prior fiscal year and the current fiscal year. This definition is similar to Chay and Suh (2009) with the only

PAGE 28

28 difference being that I annualize the standard deviation and multiply by 100 to convert from a decimal to percentage. This is done to bring the scaling closer to that of the VIX variable to allow for closer comparison of coefficients. This definition is utilized as a proxy for firm level cashflow volatility, despite the fact that it also contains some portio n of market level volatility in it, to remain consistent with prior literature. However, by containing market level volatility within the proxy for cashflow volatility I am biasing against my key independent variable VIX as some of its power is being abs orbed by ReturnVolatility The variable ReturnVolatility also has the issue that it can proxy for start up firms and distressed firms which are less likely to pay dividends and more likely to decrease/eliminate dividends if they had paid them in the prior fiscal year. My main variable of interest is the variable VIX which is defined as the average value of the Chicago Board Options Exchange Volatility Index (VIX) for the first nine months of the firms fiscal year. The VIX represents the implied volatilit y of index options of the S&P 500. It is commonly referred to as the markets fear index. Using the VIX in my specifications allows me to measure the impact of market level volatility as well as the impact of the markets anticipation for the future di rection of the market. As indicated by the fear index moniker, the VIX can also be utilized to capture some of the behavioral impact at the market level. When the VIX increases significantly, it generally indicates that the market has begun a downturn or the market believes tough times are ahead; or a combination of both. As Lintner (1956) points out, firms are hesitant to make major changes to prior dividend policies. Likewise, as Brav et al (2005) show using survey evidence, firms tend to make divi dend decisions using the prior fiscal years dividend per share as a

PAGE 29

29 base. Therefore, it would seem logical that the VIX would have a significant impact on the year to year decision of whether or not to change the payout policy (increase, decrease, eliminate, initiate, or maintain). Firms will consider the prior dividend per share level and then make changes relative to this level based upon firm level and market level information. A firm is likely to decrease the per share dividend only if current free cash flow and/or future cash levels are threatened due to a change in either internal or external indicators. As a result, large changes in the VIX may be one of the driving factors causing firms with steady dividends to decrease/eliminate their prior div idend per share, or for a firm with a steadily increasing dividend per share over time to cease the increase and maintain the current level. However, a market wide indicator such as the VIX should be much less likely to have a major impact as a general i ndicator of which firms pay dividends verse which firms dont pay. Given this is a relatively stable choice (firms generally either pay over time or dont) it wouldnt be expected that a variable that is not firm specific would have a significant impact. In fact, the most likely impact the VIX should have on the tests of payers verse nonpayers is through increases (decreases) in the number of firms eliminating (initiating) dividends or significantly decreasing (increasing) dividends when the VIX is high, and vice versa for when the VIX is low. Since changes in the pure level of dividends per share will not be caught through tests only dealing with the choice to pay or not to pay, the impact of the VIX should be significantly less important for the choice to pay a dividend. The final volatility variable utilized is the dispersion of analyst opinion. AnalystDispersion is measured as the dispersion of opinion amongst analyst forecasts

PAGE 30

30 scaled by the mean monthly price of the firms stock. For AnalystDispers ion to be a nonmissing value I require a minimum of three analysts to have covered the stock during at least four months of the prior fiscal year. Analyst dispersion is often utilized as a measure of uncertainty about the firms future outlook. If analy sts, (who are generally regarded as one of the most, if not the most, informed investors regarding the stocks they cover) have large disagreements regarding potential future earnings, then it is likely that upper level management also has the same feeling when looking at future earnings of the firm. Therefore I utilize dispersion as a way to capture firm level volatility in a forwardlooking environment. The prior expectation is that firms that have more volatile outlooks on future earnings should be more likely to have a conservative approach towards dividend policy. This should result in a higher propensity to decrease dividends per share and a lower propensity to increase. Unlike the VIX measure, analyst dispersion is a firm level characteristic and is likely to be more stable over time (firms with high dispersion of analyst opinion during one fiscal year are more likely to have had high dispersion in prior years) so it should also impact the payer/nonpayer choice significantly as well. However, usi ng analyst dispersion is not without its issues. As Das, Levine and Sivaramakrishnan (1998), Lim (2001) and others have shown, analyst forecasts are subject to their own biases. These biases need to be taken into consideration when evaluating the impact o f AnalystDispersion Also, due to the requirements of having at least three analysts covering the stock for at least four months of the fiscal year, AnalystDispersion has a significant impact on the size of the sample which can be utilized. This effect i s likely to bias the sample towards a higher percentage of dividend

PAGE 31

31 paying stocks as large well established firms which have more analyst coverage and will be more likely to pay dividends. Therefore I run each regression with and without AnalystDispersion so as to not affect the interpretation of other variables due to a restricted sample. For each of the three volatility measures ( ReturnVolatility VIX and AnalystDispersion ) I employ interaction terms allowing the potential to identify the differing impa ct each measure of volatility has depending on the firms level of cash flow This is an important addition to my study and, at least to the extent of my knowledge, is the first time this approach has been utilized in the payout policy literature. It may be the case that internal, firm level volatility or external, market level volatility impacts a firms payout decision differently if they have relatively low cash flow than it would if they currently had a relatively high level of cash flow. The three volatility measures are interacted with cash flow quartiles which are dummy v ariables that place each firm into a quartile based upon its relative level of cash flow for the current fiscal year. The cash flow variable used for the quartiles is defined as cash flow divided by total assets where cash flow is measured as earnings bef ore interest, taxes and depreciation (EBITDA in Compustat). Quartile 1 represents the 25% of firms with the lowest levels of cash flow over assets and Quartile 4 represents the 25% of firms with the highest levels of cash flow over assets. The prior expec tation is that the level of cash flow will have a significant impact on the effect of the external, market level volatility measure VIX The intuition for this expectation is that firms that are doing relatively well during a tough market may make minor adjustments to payout policy (such as simply maintaining their dividend per share

PAGE 32

32 instead of giving a slight increase) as a reaction to the tough economic environment, but are much less likely to make major changes (such as decreasing or eliminating their d ividend per share) than firms that are being hit relatively hard by the poor market. In fact, it may be that firms in the upper quartile of cash flow during high VIX periods actually go as far as to take advantage of the market downturn and increase dividends or repurchases during these tough economic periods. Similar to the Catering Theory argument made by Baker and Wurgler (2004), this would imply that firms are looking for times in which they can take advantage of their relative strong position as a ch ance to cater to investor preferences for dividends during tight markets. Firms that would be likely to fall in to this category would be firms with high cash flow levels and more stable earnings. For the firm level volatility measures ( ReturnVolatilit y a nd AnalystDispersion) the prior expectation is that there should be no significant difference between firms with relatively low cash flow and relatively high cash flow for the current fiscal year. If these firms are concerned about avoiding constant chang es to the dividend level (as Brav et al (2005) show most firms generally are), then firms with high volatility should not be eager to increase or initiate dividends just because they have one fiscal year with high earnings. Instead they should be concerne d that, due to their high volatility of earnings, they may be forced to turn around the next fiscal year and pay the penalty of decreasing/eliminating this newly established dividend level and should therefore avoid the change altogether. However, it coul d be that firms with high internal volatility are less sensitive to the dividend stickiness argument of Lintner (1956) and therefore frequently adjust dividend per share levels, and even potentially whether they pay a

PAGE 33

3 3 dividend or not, to fit their current cash flow levels. If this were the case, the results should indicate that firms in cash flow quartile 4 specifically, and possibly those in cash flow quartile 3 as well, should be more (less) likely to increase (decrease) their dividends even if they have high internal volatility. It may also be the case that the same firms in cash flow quartiles 3 and 4 may be more likely to be dividend payers regardless of internal volatility. Tax variables. I utilize two tax specific variables to show the effect tax law changes have on payout policy choices. The first tax variable is the ChangeDivTaxRate which is simply the one year change in the top United States dividend tax rate. The prior expectation for ChangeDivTaxRate is that when the tax rate increases firm may become less likely to pay dividends (or pay as high of a level of dividends per share). On the other hand, when tax rates are decreased dividends should become more desirable for investors and firms may look to increase/initiate dividends. The other tax related variable I utilize is RepatTaxCutDummy which is set equal to one for firms that have positive foreign income (Compustat Item #273) during fiscal years 2004, 2005 or 2006. These years included repatriation tax cuts under the American Job Creation Act. It is likely that firms will be more inclined to initiate or increase (decrease or eliminate) dividends and repurchases in years when the dividend tax rate decreases (increases). Similarly, firms with positive foreign income are likely to be more (less) inclined to initiate or increase (eliminate or decrease) dividends since they are the firms that are subjected to the tax break. It may also be the case the additional foreign income that is repatriated will be utilized for repurchases rather than

PAGE 34

34 dividends. Repurchases are generally viewed as more flexible and therefore may be more appropriate for a onetime increase in payouts. Additional firm level controls Other firm level controls which are utilized include CashFlow/Assets, LagNetDebt/Ass ets, NegRetainedEarn and LagReturn. CashFlow/Assets is defined as cash flow divided by total assets where cash flow is measured as earnings before interest, taxes and depreciation (EBITDA in Compustat). Firms with higher levels of cash flow should be in a better position to pay out dividends than firms with lower levels of cash flow. The interaction quartiles are based on this measure. LagNetDebt/Assets is calculated by dividing the value of net debt in the firms prior fiscal year by the value of total assets (Compustat Item #6) in the firms current fiscal year. The value of net debt is set equal to long term debt (Compustat Item #9) plus debt in current liabilities (Compustat Item #34) minus cash and short term investments (Compustat Item #1). Firms with large values of net debt from the prior fiscal year should be less likely to initiate or increase their dividends as a substantial portion of free cash may be needed to pay down portions of the debt. NegRetainedEarn is a dummy variable which is set equal to one if the value of the firms retained earnings (Compustat Item #36) is negative; and 0 otherwise. If a firm has negative retained earnings it is very unlikely that they are paying a dividend. In many cases it may actually be a covenant violati on to pay a dividend with negative retained earnings. LagReturn is the oneyear stock return for the prior fiscal year. Firms that have had substantial rundowns in stock price should be more likely to decrease the level of

PAGE 35

35 their dividends. The cause for this can be twofold. If the firms stock price is falling because of a decrease in earnings the firm is likely forced to cut its dividend. However, if it is due to a changing environment around the company and its future outlook the firm may still pre fer to decrease the dividend to keep the dividend yield from increasing substantially while the likelihood of maintaining the new higher dividend yield becomes low. Summary Statistics Summary statistics for my sample are given in Table 2 2 through Table 24 Table 2 1 represents the entire sample. Slightly over one quarter (26.7%) of all firms in my sample pay a dividend. This number increases to 34.5% when looking at firms with positive repurchases throughout the fiscal year. Taking into consideration both dividend paying and nondividend paying firms, the average dividend per share is approximately 16 cents per share. In Table 23 I break the sample down into dividend paying stocks and nondividend paying stocks. This allows comparisons of common characteristics to each group. In general dividend paying firms are signi ficantly larger, more than doubling the average NYSE percentile classification of nondividend paying firms. Dividend paying firms are likely to have a lower market to book ratio, more debt, higher earnings and cash flow and fewer investment opportunities (as proxied by dA/A in Fama and French (2001)). Firms that pay dividends are also more likely to have positive share repurchases. Consistent with the fact that dividend paying firms are less likely to be high volatility growth firms, lag returns for div idend paying stocks are lower than for firms not paying dividends. Very few firms with negative retained earnings level pay

PAGE 36

36 dividends. The fact that many covenants are likely to restrict firms from paying dividends if they have negative retained earnings is a probable cause for this finding. Table 24 looks at firms with positive share repurchases during the fiscal year relative to firms with no share repurchases. Repurchasing firms share many of the same characteristics in general as dividend paying fir ms. They are likely to be larger, have lower market to book, higher earnings and cash flow and less investment opportunities. Just as dividend paying firms were more likely to also be repurchasers, repurchasers are more likely to also be dividend payers. It is also true that repurchasers are likely to have lower returns from the prior fiscal year. Results Payout Policy Dividends Choice to pay dividends or not to pay dividends The main empirical question of this study deals with the factors that affect the payout policy choices of firms, particularly the effect of external, market level volatility. Prior literature on payout policy choice has often focused on testing this through the simple choice of paying or not paying a dividend.10 The first empiric al test I employ is to measure the effect my explanatory variables have on the probability that a firm pays a dividend. I utilize a logit model to accomplish this and the results are reported in Table 2 5 However, to best see the impact of the volatilit y measures I must first look at a linear combination of the coefficient for the base volatility variable (which represents the noninteracted lowest cash flow quartile, CFQuart1) and the coefficient on the interacted term for cash flow 10 For example, Fama and French (2001) look at the propensity to pay dividends and Baker and Wurgler (2004) look at the propensity to pay assuming you were a prior payer (PTC), propensity to pay assuming you were not a prior payer (PTI), and propensity to pay assuming you were not prior in the sample (PTL).

PAGE 37

37 quartiles 2 through 4. The results to tests on the linear combination of coefficients are reported in Table 26 Most of the firm level coefficients come through as would be expected given the summary statistics in Table 2 3 Firms are more likely to pay a dividend if they are larger and have lower market to book, higher earnings and cash flow, and have less investment opportunities (as proxied by dA/A). I also find that having negative retained earnings substantially decreases the probability of paying a dividend. As me ntioned in the Data section, this should be expected due to the fact that many covenants are likely to restrict dividend payment when retained earnings are negative. Firms are less likely to pay a dividend if they have had a rundown in share price, as ev ident by the significantly negative coefficient on lag return. Turning attention to the tax variables, Table 2 5 shows that changes in the dividend tax rate do not have a statistically significant impact on whether or not a firm pays a dividend. However, during the fiscal years associated with repatriation tax cuts due to the American Job Creation Act, it appears t hat firms with positive foreign income are actually significantly less likely to pay a dividend. This goes against the prior that firms would be more likely to take advantage of being able to repatriate their foreign income with lower repatriation taxes and then utilize this extra domestic free cash flow to payout as a dividend. On the other hand, this may simply show that firms arent willing to initiate a new payout policy over something as short term as the American Job Creation Act, but may still be w illing to change preestablished levels of dividends. Table 26 is the best way to look at the impact and significance of the key variables of interest dealing with volatility. It appears that market level volatility (proxied

PAGE 38

38 by VIX) only has an impact on low cashflow firms indicating that they are less likely to pay a dividend when the VIX is high. Internal firm level volatility, as measured by ReturnVolatility demonstrates that firms with high volatility of cash flow are not likely to pay dividends regardless of the relative level of cash flow during the current fiscal year. The variable AnalystDispersion comes through in a somewhat unexpected manner. The on ly quartile that shows as significant is quartile 3 with a statistically significant increase in the probability of paying a dividend. This is counter intuitive as the prior would be that firms with high expected volatility of future earnings would be les s likely to pay a dividend. However, this variable appears to not have strong predictive power given the nonconsistent impact it has on the regression across the quartiles. Finally, to allow comparison of the direct effect a one standard deviation change in each variable has on the probability of paying a dividend, I run marginal effects.11 The marginal effects give the impact of a one unit change from the sample means. By multiplying these by one standard deviation for the given variable, I can identify the impact a one standard deviation change in the specific variable has on the probability a firm pays a dividend. Some of the most significant variables on the probability to pay a dividend are ReturnVolatility (ranging from 15.9% for firms in cash flow quartile 1 to 20.9% for firms in cash flow quartile 4) NegativeRetainedEarn ( 12.6%) CashFlow/Asset (8.5%) NYP (7.1%) MtoB ( 4.8%) and dA/A ( 4.8%). To demonstrate how to interpret these results, a one standard deviation increase (decrease) in internal volatility ( ReturnVolatility ) results in approximately a 15.9% decrease (increase) in the probability a firm in quartile 1 pays a dividend. Similar interpretation can be utilized for 11 In order to avoid an excessive number of tables, the marginal effects tables are not reported. Full tables are available by contacting the author.

PAGE 39

39 each of the other variables. However, since NegativeRetainedEarn ings is a dummy variable the interpretation is that firms with negative levels of retained earnings are 12.6% less likely to pay a dividend. Choice to maintain, decrease/eliminate, or increase dividends per share for prior dividend payers Brav et al (2005) survey 166 financial executives of firms that pay dividends. Of these 166, 93.8% state that they consciously try to avoid reducing the level of dividends per share. Therefore, it may be of more interest to look at what factors affect the decision to ch ange dividend policy at a given point of time, rather than just the factors that affect whether a firm pays or doesnt pay a dividend. The sample is thus broken down into firms that paid a dividend in the prior fiscal year and firms that did not pay a div idend in the prior fiscal year. For the firms that did pay a dividend in the prior fiscal year their options become to maintain the current dividend per share (with a 5% buffer), decrease the dividend per share at least 5%, or increase the dividend per sh are at least 5%. Table 2 7 and Table 28 look at firms that previously paid a dividend in the prior fiscal year. Table 27 represents the multinomial logit where the base case is maintaining the current dividend level (+/ 5%). Table 28 is again the lin ear combination of coefficients for the volatility measures. Looking at the standardized marginal effects of a one standard deviation change, it is evident that firms in general look to either maintain the current dividend per share level (48. 9% at the sample means) or increase the dividend per share (42. 4 % at the sample means). Therefore it is most interesting to see what causes firms that had previously paid a dividend to

PAGE 40

40 become more likely to decrease or eliminate their dividend per share or become less likely to increase their dividend per share. First looking at the firm characteristics shows that firms with smaller (larger) amounts of both earnings and cash flow are significantly more likely to decrease (increase) their dividends. Larger market t o book ratios tend to lead to more changes in both the positive and negative direction for dividends. Firms that have high asset growth are more likely to increase dividends and less likely to decrease dividends. This would appear to go against the prior for this term if in fact dA/A does proxy for investment opportunities as Fama and French (2001) claim. One would likely argue that firms with more investment opportunities would be less likely to pay a higher dividend, and would be instead more likely to try and retain the capital to invest within the company. More (less) debt leads to a lower (higher) probability of increasing dividends per share. Firms with negative retained earnings are much less likely to increase dividends and much more likely to d ecrease or eliminate as would be expected based on covenants. Finally, firms with strong runups in price are more likely to be increasing their dividend, while firms with large price drops are more likely to decrease. This should be the case for several reasons including that firms may be in poor (strong) financial health as the stock price falls (rises) and may also be looking to keep the dividend yield from changing substantially as the price moves significantly in one direction. Changes in the dividen d tax rate appear to only be a significant factor when it comes to decreasing the dividend per share. When the dividend tax increases (decreases) firms become significantly more (less) likely to decrease (increase) their

PAGE 41

41 dividend level. However, the repatriation tax appears to show a significant effect only for increasing the dividend. When firms with positive foreign income receive a tax break on repatriating foreign profits back to the United States they take the opportunity to increase dividends to return some of the foreign income back to the investor. The linear combinations (Table 28 ) again provide the clearer way to interpret the key volatility measures and the differing impact based on the firms relative cash flow level. Firm level volatility, represented by ReturnVolatility shows that firms with high internal volatility are more (less) likely to decrease (increase) their dividend. This appears to hold true at all cash flow quartile levels. However, in a couple of cases statistical significan ce is lost. For the VIX it appears that the firms relative cash flow position causes the VIX to have a differing effect. Firms with high cash flow are relatively unaffected by volatile market conditions and appear no more likely to decrease dividends t han they would be otherwise.12 Howeve r, for firms with low cash flow high market volatility does have a significant impact resulting in a significantly higher probability of decreasing their dividend. On the other hand, a high level of VIX does affect fir ms that would have likely increased dividends otherwise regardless of relative cash flow. Though the impact is stronger in general on firms with low cash flow, all levels of firms appear to become more conservative about increasing dividends when the VIX measure is higher. This shows that firms are prone to try and maintain extra free cash flow to ensure that they 12 It should be noted that another explanation exists besides the Lint ner (1956) dividend stickiness argument for why firms with high enough cash flow levels to be able to maintain their current dividend level would prefer not to adjust their dividend level down. Fuller and Goldstein (2011) show that dividendpaying stock s have higher average returns during declining markets than their nondividendpaying counterparts. Therefore, firms may prefer to try to maintain (or even increase) their dividend level if their cash level is high enough to allow, even ignoring the penal ty that is often associated with decreasing.

PAGE 42

42 have enough cash to make it through until the market stabilize s. These periods are usually associated with a tightening of the credit markets and having internal cash should be more valuable during these times. With regard to volatility through the analyst dispersion measure it appears that firms with high expected volatility of future cash flow are more likely to decrease dividends, regardless of relative levels of cash flow On the other hand, firms with higher levels of cash flow and high analyst dispersion are significantly less likely to increase their dividend than firms with lower levels of cash flow and high analyst dispersion. This ag ain is slightly counterintuitive and relatively unexplainable. A firm with relatively low levels of cash coming in and relatively high volatility in future expected earnings would seem to be extremely unlikely to increase its per share dividend. However, they appear to be more likely to increase their dividend than their counterparts with higher cash flow levels. As with the choice of whether to pay or not, I again run marginal effects and look at the impact of a one standard deviation change from the sam ple means Given there are now three choices (maintain dividend, decrease/eliminate dividend, or increase dividend) this allows comparisons of how probabilities are transferred based on the variables. One of the more interesting observations from this oc curs when firms have negative retained earnings. As has already been discussed, covenants may require firms to decrease/eliminate dividends during this circumstance; or may at least prohibit them from increasing dividends from the prior level. This shows quite significantly as firms become 11. 0 % less likely to increase their dividend, 7.8% more likely to decrease, and 3. 2 % more likely to maintain.

PAGE 43

43 Firms with low cash flow and high internal volatility (as measured by ReturnVolatility ) prove to be more likely to change their policy in general. A one standard deviation increase in ReturnVolatility for firms in the lowest cash flow quartile results in a 18.7 % lower probability of the firm maintaining its prior dividend per share, a 8.3 % higher probability of decreasing and a 1 0 .4% higher probability of increasing. Turning to the VIX measure shows that a change in the VIX has a very economically significant impact on the firm dividend choice. Just a one standard deviation increase in the VIX (the recent crisi s resulted in substantially more than a two standard deviation change for comparison sake) makes a firm in the lowest cash flow quartile 9.4% less likely to increase its dividend per share on average. Many firms then simply maintain their current dividend level instead (firms are 3.1 % more likely to maintain) while others are actually forced to decrease (firms are 6.3 % more likely to decrease). Given the base probability of decreasing/eliminating a prior dividend level is only 8. 7 % it becomes quite evident how a several standard deviation change, as with the recent financial crisis, can cause a substantial rise in the number of firms decreasing/eliminating prior dividends (as can be seen in Figure 2 1 and the press release quotes related to recent dividend announcements in the Introduction Section). Choice to maintain no dividend or initiate a dividend for prior dividend nonpayers Table 2 9 and Table 210 look at the portion of the sample that did not pay a dividend in the prior fiscal year. For these firms the dividend choice is simply whether or not to initiate a dividend. The firm characteristics of firms that initiate dividends are similar to the fir m characteristics that distinguished payers from nonpayers in Table 2 5 This should be the case in general as these firms are now joining the group of

PAGE 44

44 payers that were being described. It appears that firms are much more willing to initiate a dividend only when cash flow levels are high and debt levels are low. This would signify times when the cash position of the firm is most positive allowing them the ability to return some free cash to investors. It also appears that firms are timing their initiati ons when looking at the external volatility (VIX) measure. Firms are quite unwilling to initiate a dividend during a poor economic environment. Given that it is commonly recognized that there is a penalty associated with decreasing/eliminating a prior es tablished dividend policy firms are likely to wait for market conditions that they feel comfortable are conducive to maintaining an acceptable level of earnings to continue a newly initiated dividend policy.13 Therefore when the VIX is high, representing a very unstable (and likely down) market, firms tend not to initiate a dividend. This is particularly true for the firms with lower levels of cash flow during high VIX periods. The firm specific volatility measure also is associated with a lower likelihood of dividend initiation. Payout Policy Repurchases Choice to repurchase or not repurchase shares. Skinner (2008) demonstrates that repurchases are increasingly being used both in addition to and in place of dividends as a means to pay out earnings to investors. Thus it is important to also consider the impact internal and external volatility, taxes, and the rest of my independent variables have on repurchases and not just on dividends. Table 2 1 1 and Table 21 2 look at the choice to either repurchas e or not repurchase shares in a logit regression similar to Table 2 5 and Table 25 for dividends. 13 See Brav et al (2005) for survey evidence that upper level management recognizes the penalty associated with decreasing/eliminating an established dividend.

PAGE 45

45 The factors that affect the choice to repurchase shares are, for the most part, quite similar to the factors that affect the choice to pay dividends. Table 2 1 1 shows that larger firms with lower market to book ratios, higher earnings and cash flow, and less investment opportunities are more likely to pay dividends. Having a large positive lag return and negative retained earnings also both make a firm less likely to repurchase as they did with dividends. However, the tax and volatility measures appear to have very different impacts on the repurchase choice than they did for the dividend choice. The change in dividend taxes works in the opposite direction t hough it is still insignificant. The repatriation tax cut dummy on the other hand flips sign and is significantly positive for repurchasing. This finding would appear to imply that when dividends become less attractive to investors firms turn more toward repurchases. A likely explanation for this is that repurchases are often considered by management to be more flexible than dividends. Therefore, management can utilize a onetime increase or initiation in repurchases to distribute the onetime additional cash flow from the repatriation tax break to investors without facing the dividend stickiness issues that would arise from an increase or initiation of dividends.14 The volatility measures all change to some degree when considering the repurchase decisi on instead of the dividend decision. In the linear combinations in Table 2 6 the VIX only mattered for the lowest cash flow firms making them less likely to pay a dividend. However, for the repurchase decision, Table 2 1 2 shows that the VIX only matters for the higher cash flow firms (becoming increasingly significant as cash 14 Dividend stickiness refers to the argument made in Lintner (1956) which claims that firms are hesitant to increase or initiate dividends without a strong confidence that the new dividend level can be maintained in order to avoid the penalty associated with decreasing or eliminating the dividend in subsequent periods.

PAGE 46

46 flow increases) and makes them more likely to repurchase. The explanation for this would appear to be that firms with high cash flow are actually taking advantage of a beat down stock price (which should be more likely during a high VIX period) to purchase undervalued shares. Firms with high internal volatility ( ReturnVolatility ) and high expected future internal volatility ( AnalystDispersion) are, in general, less likely to repurchase shares. The most likely explanation for this would appear to be that these firms are more reluctant to repurchase when their firm specific volatility increases in order to maintain a buffer level of cash within the firm. The effect of having high fir m volatility and high cash flow comes through with very strong economic significance in the marginal effects of a one standard deviation change as well. For firms in cash flow quartile 3 or 4 a one standard deviat ion increase in volatility of returns results in a decrease in the likelihood of repurchasing of 7.8 % or 9.4 %, respectively. Given the base probability of repurchasing at the sample means is 32. 8 % these are clearly very economically significant. A one st andard deviation increase in the VIX results in firms in cash flow quartiles 3 and 4 hav ing an increased probability of repurchasing of 6. 2 % and 7.9 %, respectively. Choice to maintain, decrease/eliminate, or increase repurchasing for prior repurchasers. T able 2 1 3 and Table 21 4 show the choice to maintain, decrease/eliminate, or increase the level of share repurchases given that the firm repurchased shares during the prior fiscal year. The results in this multinomial regression are somewhat difficult to interpret as, even after increasing the range considered to be maintaining the prior repurchase level to a generous +/ 20%, a

PAGE 47

47 relatively small percentage of firms fall into the maintain category. This demonstrates that firms have an extremely large amount of fluctuation in their repurchase policy. Very few variables come through as significant in determining whether a firm increases their repurchase level. It appears that smaller firms with littleto no debt and high returns during the prior fiscal year are the most likely to be increasing their repurchase levels. Firms with small cash flow are less likely to increase repurchases when the VIX is high while firms with high cash flow are more likely to increase repurchases when their return volatility is hi gh. This is likely demonstrating that firms with cash available and volatile stock prices are trying to time their repurchasing while firms with low levels of cash are concerned with retaining cash, particularly when the market appears unstable. It appear s that the decision to decrease repurchases is largely dependent upon cash flow and earnings. Interestingly, as with repurchase increas es firms with high cash flow and high return volatility are also more likely to decrease repurchases. Again, this is likely due to them utilizing their cash and volatility of stock price to try and time their repurchases. Therefore they are more likely to both increase and decrease repurchases as they look for low stock price times to repurchase shares. Given that repurchases are highly unstable over time, it is not clear that the change in repurchase regressions are all that meaningful compared to the decision to simply repurchase or the decision to repurchase after not doing so in the prior year. Choice to maintain no repurchasing or initiate a repurchase for prior nonrepurchasers. Table 2 1 5 and Table 216 look at the choice to initiate repurchases given the firm did not repurchase shares during the prior fiscal year. Firms

PAGE 48

48 with low market to book, a low level of investment opportunities, high cash flow, little to no debt and positive retained earnings are the most likely to initiate repurchases. These would all be things that in general would describe a firm with money available to return to investors and not a lot to use it for internally. It would make sense that the firms that would fit in that category would be the ones most likely to utilize a short term method to increase the portion of the cash being returned to investors. If the firm felt that a signifi cant portion of the cash that was available would continue to be available in future years, they may also increase/initiate dividends. However, repurchases are a way to get some of the cash to investors without having to worry as much about being able to maintain the payout in the future. Firms with high cash flow and high return volatility are actually less likely to initiate repurchases. This seems to go a little bit against intuition as it may seem that firms that have experienced high internal volatility but had significant cash would be more likely to use repurchases than dividends. On the other hand, these firms may actually be doing neither and are instead maintaining the cash due to the high internal volatility. In regards to external volatility, it appears that when the VIX is high, firms with high cash flow are more likely to initiate repurchases. This can again likely be attributed to trying to time repurchases. Firms are utilizing this excess cash to buy back underpriced shares when t he market fear is increasing. GARCH Robustness Check One potential concern that arises from my original sample is that the results are being carried by one or two periods of significant change. My dataset only looks at the years 1990 to 2009 due to the fact that VIX is only calculated back to 1990. However, with only tw enty years of data the recent financial crisis, and to a lesser extent the

PAGE 49

49 technology bubble of the early 2000s, may make up too significant a portion of my sample. Since the financial crisis saw such large swings in the VIX, just one or two years of data may be influencing my findings quite significantly. To offset this concern I utilize a GARCH estimate of the VIX which allows the dataset to expand back to 1962. I then utilize this new dataset to test if my findings still hold over a nearly fifty year sample. Details of the calculation of the GARCH estimate, as well as all tables discussed in this section, are available in the Appendix. To first ensure that the GARCH estimate is a good proxy for the VIX I test the correlation between the two variables. I find that a simply GARCH (1,1) estimate over the entire sample period results in a correlation of approximately 0 .913. I then use the GARCH estimate and re run each table from my entire study replacing the VIX variable with the GARCH estimate. I find that my results are not significantly changed with the GARCH estimate as coefficients and magnitudes remain quite stable. Next, I rerun each logit and multinomial logit on my new full sample consisting of data from 1962 to 2009. The main findings of my s tudy regarding dividend choice remain relatively unchanged. Using the expanded dataset I still show that low cash flow firms are more likely to decrease or eliminate their dividends when market level uncertainty is high. Similarly, firms across all cash flow levels are unlikely to initiate dividends during times of high market uncertainty. It appears that the recent financial crisis is not solely responsible for my finding that firms become conservative with dividend payouts during high market uncertaint y periods given the results hold even with an expanded dataset.

PAGE 50

50 However, the results regarding repurchases do encounter some slight changes when expanding the sample. This makes some sense as the popularity of repurchasing shares has changed substantially since the early 1960s. Therefore, by adding an additional nearly thirty years of data, it may be expected that the results would change to some degree. In the extended sample tests I show that high levels of the GARCH estimate result in increased probability of repurchasing shares and increased probability of initiating a repurchase, regardless of cash flow level. This is slightly different than from the original sample, in which only high cash flow firms were more likely to repurchase or initiate a repurchase during high VIX level periods. To further ensure that my results are not being carried by the high market uncertainty period of the recent crisis, I also look at only the period 1962 to 1989. This allows me to see the differing impact of all variables prior to the VIX period of my original sample. Again, my original findings appear to stand up to this changed sample period. However, there again are some slightly different impacts for some variables with the change of sample period. While low cas h flow firms are still most impacted by periods of high market uncertainty with regards to dividend policy, the impact also appears to impact high cash flow firms more during this early sample than it did during the VIX sample. With regar ds to repurchases the 1962 to 1989 period demonstrates that high market uncertainty can be linked to an increased probability of repurchasing regardless of cash flow level. By using a GARCH estimate of VIX to extend my sample, I believe I demonstrate that my results are r obust to changing sample periods. Despite concern that the recent financial crisis may be dominating the VIX sample, it appears that this is not the case.

PAGE 51

51 Even after removing the VIX sample period from the GARCH dataset and rerunning all tables on only t he 1962 to 1989 period, the findings are mostly unchanged. However, it does appear that extending the sample period does have some impact on my repurchase choice results as firms become more likely to repurchase or initiate a repurchase during high market uncertainty periods regardless of cash flow level. Overall, I believe this extension of the dataset should ease concern over the possibility of one period of high market uncertainty having too large of an impact on my results. Conclusion In this study I show that market level volatility has an important influence on payout policy. This result holds even after controlling for firm level cash flow volatility as well as a host of other firm level characteristics from prior literature When the market becom es unstable, as proxied by high values of the Chicago Board Options Exchange Volatility Index (VIX), firms make different payout policy decisions than when the market has low volatility/uncertainty. However, I show that the effect of the volatility measures, both internal and external, are not always consistent across firms in different cash flow positions. In many cases, fir ms with low levels of cash flow react differently (in direction and/or magnitude) than firms with higher levels of cash flow. This may be expected as firms with low cash flow levels may have their hands forced by market level uncertainty, while firms with high cash flow may actually be able to take advantage of the uncertainty. In general dividend paying firms with low cash flow are significantly affected by high levels of the VIX. They became significantly more likely to decrease their dividends and significantly less likely to increase their dividends when the VIX is high. However, firms with high cash flow are relatively unaffected. On the other hand, internal

PAGE 52

52 volatility (proxied for by the volatility of returns as in Chay and Suh (2009)) tends to cause firms to be more likely to decrease dividends regardless of cash flow and has little impact on the choice to increase dividends. For nondividend paying firms, internal and external volatility influences the choice of whether or not to initiate dividends. High market volatility has a very significant negative effect on the probability of initiating whether the firm has high cash fl ow or low cash flow. Firms appear to push off the decision to initiate until after the market stabilizes. The same is true for internal volatility. Firms with high internal volatility are significantly less likely to initiate dividends regardless of rel ative cash flow. In addition to the volatility measures, tax variables are added to look at the effect of dividend tax changes as well as cuts in the repatriation taxes on dividend payouts. Firms do appear to consider tax changes when making payout decisi ons. When the top tax rate is increased (decreased) firms become more (less) likely to decrease dividends in response. I also show that firms with positive foreign income utilized the American Job Creation Acts repatriation tax relief as an opportunity to increase dividend payouts. Furthermore the repurchase choice is considered in the same framework to allow the opportunity to see how internal and external volatility affect the repurchase decision. The results seem to indicate that firms that have hig h cash flow available tend to try and time the market to allow repurchasing when prices are beaten down by a volatile market. The probability of a firm having positive share repurchases increases significantly when the VIX is high, given that they have relatively high levels of cash flow. This holds true for the probably of a firm initiating a repurchase as well. However,

PAGE 53

53 the results are less clear for the choice of the repurchase level for prior repurchasers. The impact of internal volatility on the repurchase decision is also relatively vague. To ensure that my results are not being driven by the high VIX period that occurred during the recent financial crisis, I calculate a GARCH (1,1) estimate of the VIX which allows my dataset to expand to 1962 through 2009. I then rerun each test using the entire 1962 to 2009 sample as well as the nonoverlapping 1962 to 1989 sample. In both the extended overlapping sample as well as the nonoverlapping sample, I find that my results are relatively unchanged. Thi s should demonstrate that the results are not only capturing a change relating to a onetime event, but instead a true impact on payout policy caused by market level uncertainty.

PAGE 54

54 Table 21 Variable d efinitions. Variable Definition AnalystDispersion The dispersion of opinion amongst analyst forecasts scaled by mean forecast (utilizing IBES Detail Estimates). For this variable to be included in the sample firms are required to have a minimum of 3 analysts covering their stock for a m inimum of 4 months during that fiscal year. Book Value Stockholders equity (Compustat Item #216) minus preferred stock (Compustat Item #10) plus deferred taxes and investment tax credits (Compustat Item #35) plus post retirement assets (Compustat Item #15) CashFlow/Assets Cash flow divided by total assets where cash flow is measured as earnings before interest, taxes and depreciation (EBITDA in Compustat) CFQuart2 Dummy variable that is equal to 1 if the firm falls into the second quartile (25% to 50%) for CashFlow/Assets relative to the other firms in the sample, and 0 otherwise CFQuart3 Dummy variable that is equal to 1 if the firm falls into the third quartile (50% to 75%) for CashFlow/Assets relative to the other firms in the sample, and 0 otherwise CFQuart4 Dummy variable that is equal to 1 if the firm falls into the fourth quartile (75% to 100%) for CashFlow/Assets relative to the other firms in the sample, and 0 otherwise ChangeCapGainsTax Equal to the current years top capital gains tax rate minus the prior years top capital gains tax rate ChangeDivTax Equal to the current years top dividend tax rate minus the prior years top dividend tax rate dA/A (Total assets lag(total assets)) divided by lag(total assets) DividendPerShare The ex date dividends per share for the fiscal year (Compustat Item #26) E/A (Income before extraordinary items (Compustat Item #18) plus interest and related expenses (Compustat Item #15) plus deferred incomes taxes (Compustat Item #50)) divided by total assets

PAGE 55

55 Table 21 Continued Variable Definition GARCH Estimate of the VIX using a GARCH(1,1) estimation of the volatility of logarithmic S&P 500 returns. LagNetDebt/Assets Lag(net debt) divided by total assets where net debt is equal to longterm debt (Compustat Item #9) plus debt in current liabilities (Compustat Item #34) minus cash and short term investments (Compustat Item #1) LagReturn The one year stock return for the prior fiscal year MtoB Measure of the firms market to book value defined as (total assets (Compustat Item #6) minus book value plus market equity) divided by total assets Market Equity Stocks closing price (Compustat Item #24) multiplied by the number of common shares outstanding (Compustat Item #25) NegativeRetainedEarn Dummy variable that equals 1 if the firm has negative retained earnings (Compustat Item #36) over total equity (Compustat Item #144) for the current fiscal year, and 0 otherwise NYP The firms market capitalization percentile relative to all NYSE firms (calculated in 5% intervals) RepatTaxCutDummy Dummy variable that is equal to 1 if the firm has positive foreign income (Compustat Item #273) during the American Job Creation Acts repatriation tax cut years of 2004, 2005, and 2006 ReturnVolatility The annualized standard deviation of the prior 24 monthly returns including the current fiscal year multiplied by 100 VIX Mean value of the CBOE Volatility Index for the first nine months of the firms fiscal year

PAGE 56

56 Table 2 2 Summary s tatistics for the whole sample. This table shows summary statistics for the entire sample. Data is from 1990 through 2009 Standard Deviation Percent Equal to 1 Variable Observations Mean Minimum Maximum NYP 75539 0.32 0.31 0.05 1.00 M/B 75653 1.99 1.67 0.54 10.71 dA/A 75650 0.17 0.50 0.56 2.99 E/A 69956 0.02 0.23 1.24 0.25 VIX 74556 19.80 6.00 12.13 45.69 ReturnVolatility 75568 59.41 40.32 14.66 272.19 CashFlow/Assets 75435 0.05 0.22 1.01 0.39 LagNetDebt/Assets 67890 0.02 0.35 0.89 0.66 NegativeRetainedEarn 75653 39.0% ChangeDivTax 75653 0.01 0.05 0.24 0.09 RepatTaxCutDummy 75653 4.4% AnalystDispersion 31851 0.01 0.01 0.00 0.10 LagReturn 68135 0.48 2.21 0.88 16.68 Div/Share (All) 75653 0.16 0.60 0.00 51.81 Positive NetDebt 75653 57.3% Positive Dividend 75653 26.7% Repurchaser 75653 34.5% Variable winsorized at 1% level

PAGE 57

57 Table 23 Summary statistics for dividend nonpayers verse dividend payers. This table shows summary statistics for the sample broken down into firms that did not pay a dividend vs firms that did pay a dividend during t he current fiscal year. Data is from 1990 through 2009. Summary Statistics No Dividend Summary Statistics Dividend Standard Deviation Percent Equal to 1 Standard Deviation Percent Equal to 1 Variable Obs. Mean Min Max Obs. Mean Min Max NYP 55342 0.25 0.27 0.05 1.00 20197 0.52 0.32 0.05 1.00 M/B 55453 2.11 1.91 0.52 11.99 20200 1.68 0.95 0.65 6.05 dA/A 55450 0.20 0.59 0.59 3.55 20200 0.09 0.21 0.33 1.19 E/A 50545 0.06 0.27 1.42 0.25 19411 0.07 0.06 0.18 0.25 VIX 54655 19.95 5.98 12.13 45.69 19901 19.40 6.04 12.13 45.69 ReturnVolatility 55376 69.01 45.11 18.92 316.28 20192 34.19 14.28 12.79 89.76 CashFlow/Assets 55299 0.01 0.25 1.16 0.38 20136 0.15 0.08 0.06 0.42 LagNetDebt/Assets 48193 0.02 0.37 0.92 0.68 19697 0.11 0.25 0.64 0.61 Neg RetainedEarn 55453 50.8% 20200 6.7% ChangeDivTax 55453 0.01 0.05 0.24 0.09 20200 0.00 0.05 0.24 0.09 RepatTaxCutDummy 55453 3.4% 20200 6.9% AnalystDispersion 19621 0.01 0.02 0.00 0.14 12230 0.00 0.01 0.00 0.04 LagReturn 48398 0.63 2.75 0.90 20.52 19737 0.12 0.61 0.68 4.41 Dividend Per Share 55453 0.00 0.00 0.00 0.00 20200 0.60 1.04 0.00 51.81 Positive NetDebt 55453 52.2% 20200 71.4% Positive Dividend 55453 0.0% 20200 100.0% Repurchaser 55453 27.0% 20200 54.9% Variable winsorized at 1% level

PAGE 58

58 Table 24 Summary statistics for nonrepurchasers verse repurchasers. This table shows summary statistics for the sample broken down into firms that did not repurchase shares vs firms that did repurchase shar es during the current fiscal year. Data is from 1990 through 2009. Summary Statistics No Repurchase Summary Statistics Repurchase Standard Deviation Percent Equal to 1 Standard Deviation Percent Equal to 1 Variable Obs. Mean Min Max Obs Mean Min Max NYP 49473 0.27 0.28 0.05 1.00 26066 0.42 0.33 0.05 1.00 M/B 49567 2.10 1.90 0.54 12.10 26086 1.78 1.20 0.55 7.47 dA/A 49565 0.21 0.58 0.58 3.52 26085 0.10 0.31 0.48 1.87 E/A 45879 0.05 0.27 1.39 0.24 24077 0.04 0.14 0.74 0.27 VIX 48798 19.42 5.89 12.13 45.69 25758 20.51 6.14 12.13 45.69 ReturnVolatility 49500 66.18 45.19 16.54 312.44 26068 46.87 28.40 13.35 179.82 CashFlow/Assets 49419 0.02 0.25 1.17 0.37 26016 0.12 0.14 0.54 0.43 LagNetDebt/Assets 43336 0.02 0.36 0.92 0.67 24554 0.02 0.31 0.81 0.63 Neg RetainedEarn 49567 47.9% 26086 22.1% ChangeDivTax 49567 0.00 0.05 0.24 0.09 26086 0.01 0.05 0.24 0.09 RepatTaxCutDummy 49567 3.5% 26086 6.0% AnalystDispersion 17906 0.01 0.02 0.00 0.13 13945 0.01 0.01 0.00 0.06 LagReturn 43487 0.61 2.68 0.90 20.04 24648 0.25 1.34 0.82 10.29 Dividend Per Share 49567 0.11 0.61 0.00 51.81 26086 0.26 0.57 0.00 31.00 Positive NetDebt 49567 56.8% 26086 58.3% Positive Dividend 49567 18.4% 26086 42.5% Repurchaser 49567 0.0% 26086 100.0% Variable winsorized at 1% level

PAGE 59

59 Table 2 5 L ogit r egression for the c hoice to p ay or n ot p ay a d ividend This table shows the logit regression coefficients for the firms choice to either pay or not pay a dividend. A firm is considered to have a positive dividend if the ex date dividend per share (Compustat Item #26) is positive. Standard errors are shown in parenthesis. Base choice is to not pay a dividend. Variable 1 2 NYP 2.217 *** 2.585 *** (0.21) (0.19) MtoB 0.274 *** 0.236 *** (0.03) (0.04) dA/A 0.931 *** 0.985 *** (0.07) (0.11) E/A 0.356 0.278 (0.29) (0.30) VIX 0.037 *** 0.030 (0.01) (0.01) ReturnVolatility 0.038 *** 0.042 *** (0.00) (0.00) CashFlow/Assets 3.726 *** 4.031 *** (0.51) (0.51) LagNetDebt/Assets 0.061 0.937 *** (0.12) (0.17) NegativeRetainedEarn 1.375 *** 1.292 *** (0.04) (0.09) ChangeDivTaxRate 0.109 0.837 (0.67) (0.89) RepatTaxCutDummy 0.226 0.359 ** (0.12) (0.15) AnalystDispersion 3.856 (3.42) LagReturn 0.053 *** 0.079 *** (0.01) (0.01) ***,**,* denote significance at the 1%, 5%, and 10% levels, respectively

PAGE 60

60 Table 2 5 Continued Variable 1 2 CFQuart2*VIX 0.031 *** 0.008 (0.01) (0.01) CFQuart3*VIX 0.030 *** 0.021 (0.01) (0.01) CFQuart4*VIX 0.031 *** 0.029 ** (0.01) (0.01) CFQuart2*ReturnVol 0.009 *** 0.001 (0.00) (0.01) CFQuart3*ReturnVol 0.011 *** 0.013 ** (0.00) (0.01) CFQuart4*ReturnVol 0.012 *** 0.017 *** (0.00) (0.00) CFQuart2*Disp 5.509 (3.84) CFQuart3*Disp 5.292 (4.64) CFQuart4*Disp 0.273 (7.75) Constant 2.339 *** 4.075 *** (0.37) (0.72) Industry Dummies Y Y Year Clustered S.E. Y Y Number of observations 61,680 26,640 Adjusted R2 0.393 0.406 ***,**,* denote significance at the 1%, 5%, and 10% levels, respectively

PAGE 61

61 Table 26 Linear combination for the choice to pay or not pay a dividend. This table shows the linear combination of coefficients for the interaction variables from Table 25 for the firms choice to either pay or not pay a dividend. A firm is considered to have a positive dividend if the ex date dividend per share (Compustat Item #26) is positive. Standard errors are shown in parenthesis. Base choice is to not pay a dividend. Variable 1 2 VIX 0.037 *** 0.030 *** (0.01) (0.01) VIX + CFQuart2*VIX 0.006 0.022 (0.01) (0.02) VIX + CFQuart3*VIX 0.007 0.008 (0.01) (0.02) VIX + CFQuart4*VIX 0.006 0.000 (0.01) (0.02) ReturnVolatility 0.038 *** 0.042 *** (0.00) (0.00) ReturnVolatility+ CFQuart2*ReturnVol 0.047 *** 0.043 *** (0.00) (0.00) ReturnVolatility + CFQuart3*ReturnVol 0.049 *** 0.055 *** (0.00) (0.00) ReturnVolatility + CFQuart4*ReturnVol 0.050 *** 0.059 *** (0.00) (0.00) AnalystDispersion 3.856 (3.42) AnalystDispersion + CFQuart2*Disp 1.653 (5.01) AnalystDispersion + CFQuart3*Disp 9.148 ** (4.03) AnalystDispersion + CFQuart4*Disp 3.583 (6.84) ***,**,* denote significance at the 1%, 5%, and 10% levels, respectively

PAGE 62

62 Table 2 7 Multinomial logit regression for the c h oice to m aintain or change d ividend for p rior d ividend p ayers This table shows the multinomial logit regression for the firms choice to either maintain or change their dividend given that they paid a dividend during the prior fiscal year. The firms dividend is measured using the ex date dividends per share for the fiscal year (Compustat Item #26). Standard errors are shown in parenthesis. Base choice is to maintain prior dividend level Variable Decrease Dividend Increase Dividen d Decrease Dividend Increase Dividend NYP 1.149 *** 0.609 *** 0.407 0.714 *** (0.24) (0.09) (0.21) (0.11) MtoB 0.283 *** 0.145 *** 0.364 *** 0.073 (0.07) (0.05) (0.08) (0.05) dA/A 0.447 ** 0.726 *** 0.339 0.759 *** (0.17) (0.11) (0.28) (0.15) E/A 1.760 *** 4.099 *** 0.433 4.905 *** (0.42) (0.96) (0.60) (1.25) VIX 0.038 *** 0.059 *** 0.039 ** 0.081 (0.01) (0.02) (0.02) (0.05) ReturnVolatility 0.033 *** 0.016 *** 0.011 0.016 (0.01) (0.01) (0.01) (0.02) CashFlow/Assets 1.426 2.238 ** 2.061 2.539 ** (0.88) (0.88) (1.34) (1.26) LagNetDebt/Assets 0.399 0.835 *** 0.054 1.360 *** (0.26) (0.11) (0.30) (0.12) NegativeRetainedEarn 0.596 *** 0.359 *** 0.534 *** 0.254 (0.06) (0.11) (0.11) (0.14) ChangeDivTaxRate 2.708 *** 0.211 1.532 *** 0.605 (0.42) (0.38) (0.47) (0.60) RepatTaxCutDummy 0.222 0.206 ** 0.059 0.201 (0.15) (0.10) (0.18) (0.11) AnalystDispersion 54.675 *** 4.420 (10.11) (13.01) LagReturn 0.655 *** 0.242 *** 0.561 *** 0.191 *** (0.07) (0.04) (0.09) (0.06) ***,**,* denote significance at the 1%, 5%, and 10% levels, respectively

PAGE 63

63 Table 2 7 Continued Variable Decrease Dividend Increase Dividend Decrease Dividend Increase Dividend CFQuart2*VIX 0.027 0.021 0.043 *** 0.053 (0.01) (0.01) (0.02) (0.05) CFQuart3*VIX 0.053 *** 0.038 *** 0.070 *** 0.064 (0.02) (0.01) (0.02) (0.05) CFQuart4*VIX 0.043 *** 0.036 *** 0.065 *** 0.061 (0.01) (0.01) (0.02) (0.05) CFQuart2*ReturnVol 0.002 0.023 *** 0.016 0.036 ** (0.01) (0.01) (0.01) (0.02) CFQuart3*ReturnVol 0.002 0.025 *** 0.021 ** 0.033 (0.01) (0.01) (0.01) (0.02) CFQuart4*ReturnVol 0.004 0.017 *** 0.017 0.026 (0.01) (0.01) (0.01) (0.02) CFQuart2*Disp 3.369 1.725 (8.21) (16.13) CFQuart3*Disp 6.208 19.964 (8.70) (21.52) CFQuart4*Disp 13.865 28.950 (12.49) (23.51) Constant 2.210 *** 0.308 3.863 *** 0.174 (0.48) (0.37) (0.66) (0.45) Industry Dummies Y Y Y Y Year Clustered S.E. Y Y Y Y Number of observations 18,553 18,553 11,266 11,266 Adjusted R2 0.158 0.158 0.168 0.168 ***,**,* denote significance at the 1%, 5%, and 10% levels, respectively

PAGE 64

64 Table 28 Linear combination for the choice to maintain or change dividend for prior dividend payers. This table shows the linear combination of coefficients for the interaction variables from Table 27 for the firms choice to either maintain or change their dividend given that they paid a dividend during the prior fiscal year. The firms dividend is measured using the ex date dividends per share for the fiscal y ear (Compustat Item #26). Standard errors are shown in parenthesis. Base choice is to maintain prior dividend level Variable Decrease Dividend Increase Dividend Decrease Dividend Increase Dividend VIX 0.038 *** 0.059 *** 0.039 ** 0.081 (0.01) (0.02) (0.02) (0.05) VIX + CFQuart2*VIX 0.011 0.037 *** 0.004 0.027 (0.01) (0.01) (0.01) (0.02) VIX + CFQuart3*VIX 0.015 0.021 0.030 *** 0.016 (0.01) (0.01) (0.01) (0.01) VIX + CFQuart4*VIX 0.005 0.022 0.025 0.019 (0.01) (0.01) (0.02) (0.01) ReturnVolatility 0.033 *** 0.016 *** 0.011 0.016 (0.01) (0.01) (0.01) (0.02) ReturnVolatility+ CFQuart2*ReturnVol 0.031 *** 0.008 ** 0.027 *** 0.020 *** (0.00) (0.00) (0.00) (0.01) ReturnVolatility + CFQuart3*ReturnVol 0.031 *** 0.010 *** 0.032 *** 0.017 *** (0.00) (0.00) (0.00) (0.00) ReturnVolatility + CFQuart4*ReturnVol 0.029 *** 0.002 0.028 *** 0.010 ** (0.01) (0.00) (0.01) (0.01) AnalystDispersion 54.675 *** 4.420 (10.11) (13.01) AnalystDispersion + CFQuart2*Disp 51.306 *** 2.694 (8.33) (10.57) AnalystDispersion + CFQuart3*Disp 48.467 *** 24.384 (4.71) (12.80) AnalystDispersion + CFQuart4*Disp 40.810 *** 33.370 ** (9.50) (14.89) ***,**,* denote significance at the 1%, 5%, and 10% levels, respectively

PAGE 65

65 Table 2 9 Logit regression for the c hoice to m aintain or p ay a d ividend for p rior dividend nonpayers This table shows the logit regression for the firms choice to either maintain or change their dividend given that they did not pay a dividend during the prior fiscal year. The firms dividend is measured using the ex date dividends per share for the fisc al year (Compustat Item #26). Standard errors are shown in parenthesis. Base choice is to maintain no dividend. Variable 1 2 NYP 0.409 0.271 (0.22) (0.22) MtoB 0.188 *** 0.247 *** (0.02) (0.03) dA/A 0.344 *** 0.598 *** (0.05) (0.09) E/A 0.026 0.005 (0.20) (0.23) VIX 0.004 0.007 (0.01) (0.02) ReturnVolatility 0.001 0.000 (0.00) (0.00) CashFlow/Assets 1.184 *** 2.049 *** (0.33) (0.50) LagNetDebt/Assets 1.010 *** 1.076 *** (0.08) (0.15) NegativeRetainedEarn 0.359 *** 0.203 *** (0.04) (0.07) ChangeDivTaxRate 0.032 0.238 (0.81) (0.92) RepatTaxCutDummy 0.222 ** 0.254 (0.09) (0.14) AnalystDispersion 2.723 (3.05) LagReturn 0.001 0.008 (0.01) (0.01) ***,**,* denote significance at the 1%, 5%, and 10% levels, respectively

PAGE 66

66 Table 2 9 Continued Variable 1 2 CFQuart2*VIX 0.012 *** 0.019 *** (0.00) (0.01) CFQuart3*VIX 0.026 *** 0.033 *** (0.00) (0.01) CFQuart4*VIX 0.032 *** 0.033 *** (0.01) (0.01) CFQuart2*ReturnVol 0.001 0.003 (0.00) (0.00) CFQuart3*ReturnVol 0.005 *** 0.009 *** (0.00) (0.00) CFQuart4*ReturnVol 0.004 ** 0.005 ** (0.00) (0.00) CFQuart2*Disp 12.798 ** (5.90) CFQuart3*Disp 0.158 (5.29) CFQuart4*Disp 6.043 (7.97) Constant 1.936 *** 1.453 ** (0.42) (0.60) Industry Dummies Y Y Year Clustered S.E. Y Y Number of observations 39,877 15,124 Adjusted R2 0.055 0.053 ***,**,* denote significance at the 1%, 5%, and 10% levels, respectively

PAGE 67

67 Table 210. Linear combination for the choice to maintain or pay a dividend for prior dividend nonpayers This table shows the linear combination of coefficients for the inte raction variables from Table 2 9 for the firms choice to either maintain or change their dividend given that they did not pay a dividend during the prior fiscal year. The firms dividend is measured using the ex date dividends per share for the fiscal year (Compustat Item #26). Standard errors are shown in parenthesis. Base choice is to maintain no dividend. Variable 1 2 VIX 0.004 0.007 (0.01) (0.02) VIX + CFQuart2*VIX 0.051 *** 0.054 *** (0.02) (0.02) VIX + CFQuart3*VIX 0.048 *** 0.049 *** (0.01) (0.02) VIX + CFQuart4*VIX 0.054 *** 0.063 *** (0.01) (0.02) ReturnVolatility 0.001 0.000 (0.00) (0.00) ReturnVolatility+ CFQuart2*ReturnVol 0.014 ** 0.026 *** (0.01) (0.01) ReturnVolatility + CFQuart3*ReturnVol 0.010 *** 0.027 *** (0.00) (0.01) ReturnVolatility + CFQuart4*ReturnVol 0.004 0.016 ** (0.00) (0.01) AnalystDispersion 2.723 (3.05) AnalystDispersion + CFQuart2*Disp 1.528 (11.42) AnalystDispersion + CFQuart3*Disp 5.002 (9.77) AnalystDispersion + CFQuart4*Disp 4.785 (8.57) ***,**,* denote significance at the 1%, 5%, and 10% levels, respectively

PAGE 68

68 Table 2 1 1 Logit r egression for the c hoice to r epurchase shares or n ot r epurchase shares This table shows the logit regression and linear combination for interaction coefficients for the firms choice to either repurchase shares or not repurchase shares. A firm is considered to be a repurchaser if total repurchases are positive, where total p urchases are defined as purchase of common and preferred stock (Compustat Item #115) plus the minimum of 0 and (preferred stock / redemption value (Compustat Item #56) minus lag(preferred stock / redemption value). Standard errors are shown in parenthesis Base choice is to not repurchase shares. Variable 1 2 NYP 0.841 *** 0.809 *** (0.22) (0.20) MtoB 0.151 *** 0.186 *** (0.02) (0.02) dA/A 0.619 *** 0.828 *** (0.06) (0.09) E/A 0.097 0.231 (0.21) (0.25) VIX 0.004 0.000 (0.01) (0.02) ReturnVolatility 0.002 ** 0.004 (0.00) (0.00) CashFlow/Assets 2.014 *** 3.660 *** (0.29) (0.48) LagNetDebt/Assets 1.058 *** 1.009 *** (0.07) (0.12) NegativeRetainedEarn 0.484 *** 0.325 *** (0.04) (0.06) ChangeDivTaxRate 0.331 0.269 (0.80) (1.06) RepatTaxCutDummy 0.337 ** 0.353 (0.15) (0.18) AnalystDispersion 1.565 (2.13) LagReturn 0.026 *** 0.038 *** (0.01) (0.01) ***,**,* denote significance at the 1%, 5%, and 10% levels, respectively

PAGE 69

69 Table 2 1 1 Continued Variable 1 2 CFQuart2*VIX 0.021 *** 0.018 (0.00) (0.01) CFQuart3*VIX 0.042 *** 0.039 *** (0.00) (0.01) CFQuart4*VIX 0.055 *** 0.056 *** (0.01) (0.01) CFQuart2*ReturnVol 0.005 *** 0.006 ** (0.00) (0.00) CFQuart3*ReturnVol 0.013 *** 0.017 *** (0.00) (0.00) CFQuart4*ReturnVol 0.015 *** 0.020 *** (0.00) (0.00) CFQuart2*Disp 8.148 (4.24) CFQuart3*Disp 2.258 (4.32) CFQuart4*Disp 20.353 ** (8.00) Constant 1.140 *** 0.596 (0.39) (0.46) Industry Dummies Y Y Year Clustered S.E. Y Y Number of observations 61,823 26,746 Adjusted R2 0.130 0.123 ***,**,* denote significance at the 1%, 5%, and 10% levels, respectively

PAGE 70

70 Table 21 2 Linear combination for the choice to repurchase shares or not repurchase shares. This table shows the linear combination of coefficients for the interaction variables from Table 21 1 for the firms choice to either repurchase shares or not repurchase shares. A firm is considered to be a repurchaser if total repurchases are positive, where total purchases are defined as purchase of common and preferred stock (Compustat Item #115) plus the minimum of 0 and (preferred stock / redemption value (Compustat Item #56) minus lag(preferred stock / redemption value). Standard errors are shown in parenthesis. Base choice is to not repurchase shares. Variable 1 2 VIX 0.004 0.000 (0.01) (0.02) VIX + CFQuart2*VIX 0.025 0.018 (0.02) (0.02) VIX + CFQuart3*VIX 0.047 *** 0.039 ** (0.01) (0.02) VIX + CFQuart4*VIX 0.060 *** 0.056 *** (0.02) (0.02) ReturnVolatility 0.002 ** 0.004 (0.00) (0.00) ReturnVolatility+ CFQuart2*ReturnVol 0.008 *** 0.010 *** (0.00) (0.00) ReturnVolatility + CFQuart3*ReturnVol 0.001 *** 0.020 *** (0.00) (0.00) ReturnVolatility + CFQuart4*ReturnVol 0.002 *** 0.024 *** ( 0.02) (0.00) AnalystDispersion 1.565 (2.13) AnalystDispersion + CFQuart2*Disp 9.713 *** (2.73) AnalystDispersion + CFQuart3*Disp 2.734 (3.44) AnalystDispersion + CFQuart4*Disp 21.918 *** (7.66) ***,**,* denote significance at the 1%, 5%, and 10% levels, respectively

PAGE 71

71 Table 2 1 3 Multinomial logit regression for the c hoice to m aintain or change r epurchase l evel for p rior r epurchasers This table shows the multinomial logit regression and linear combination for interaction coefficients for the firms choice to either maintain or change their repurchase level given that they repurchased stock during the prior fiscal year. The firms repurchase level is defined as purchase of common and preferred stock (Compustat Item #115) plus the minimum of 0 and (preferred stock / redemption value (Compustat Item #56) minus lag(preferred stock / redemption value). Standard errors are shown in parenthesis. Base choice is to maintain prior repurchase level. Variable Decrease Repurchase Increase Repurchase Decrease Repurchase Increase Repurchase NYP 0.615 *** 0.326 ** 0.502 ** 0.314 ** (0.18) (0.13) (0.23) (0.16) MtoB 0.003 0.038 0.021 0.060 (0.03) (0.04) (0.03) (0.05) dA/A 0.667 *** 0.064 0.854 *** 0.051 (0.23) (0.22) (0.29) (0.31) E/A 1.248 *** 0.500 1.487 *** 0.174 (0.40) (0.48) (0.40) (0.48) VIX 0.037 *** 0.001 0.044 *** 0.010 (0.01) (0.01) (0.02) (0.01) ReturnVolatility 0.006 ** 0.002 0.006 0.007 (0.00) (0.00) (0.01) (0.00) CashFlow/Assets 1.742 *** 0.108 2.143 *** 0.307 (0.52) (0.60) (0.79) (0.89) LagNetDebt/Assets 0.623 *** 0.288 ** 0.491 *** 0.552 *** (0.10) (0.14) (0.15) (0.17) NegativeRetainedEarn 0.041 0.063 0.194 0.071 (0.10) (0.09) (0.15) (0.16) ChangeDivTaxRate 1.572 0.461 ** 1.899 ** 1.049 *** (0.94) (0.23) (0.89) (0.29) RepatTaxCutDummy 0.361 ** 0.012 0.271 0.008 (0.14) (0.10) (0.14) (0.13) AnalystDispersion 6.180 22.894 ** (7.80) (11.61) LagReturn 0.011 0.071 *** 0.008 0.094 ** (0.04) (0.03) (0.05) (0.05) ***,**,* denote significance at the 1%, 5%, and 10% levels, respectively

PAGE 72

72 Table 2 1 3 Continued Variable Decrease Repurchase Increase Repurchase Decrease Repurchase Increase Repurchase CFQuart2*VIX 0.009 0.004 0.001 0.007 (0.01) (0.01) (0.02) (0.02) CFQuart3*VIX 0.016 ** 0.001 0.010 0.000 (0.01) (0.01) (0.02) (0.02) CFQuart4*VIX 0.047 *** 0.023 *** 0.042 ** 0.022 (0.01) (0.01) (0.02) (0.02) CFQuart2*ReturnVol 0.004 0.001 0.001 0.009 ** (0.00) (0.00) (0.01) (0.00) CFQuart3*ReturnVol 0.004 0.001 0.005 0.002 (0.00) (0.00) (0.01) (0.01) CFQuart4*ReturnVol 0.020 *** 0.012 *** 0.017 *** 0.008 (0.00) (0.00) (0.01) (0.01) CFQuart2*Disp 5.970 15.423 (15.05) (17.46) CFQuart3*Disp 1.693 12.642 (11.01) (10.55) CFQuart4*Disp 51.456 48.250 (31.47) (28.63) Constant 0.949 0.772 0.432 0.881 (0.60) (0.50) (0.85) (0.66) Industry Dummies Y Y Y Y Year Clustered S.E. Y Y Y Y Number of observations 21,944 21,944 11,616 11,616 Adjusted R2 0.057 0.057 0.065 0.065 ***,**,* denote significance at the 1%, 5%, and 10% levels, respectively

PAGE 73

73 Table 21 4 Linear combination for the choice to maintain or change repurchase level for prior repurchasers. This table shows the linear combination of coefficients for the interaction variables from Table 21 3 for the firms choice to either maintain or change their repurchase level given that they repurchased stock during the prior fiscal year. The firms repurchase level is defined as purchase of common and preferred stock (Compustat Item #115) plus the minimum of 0 and (preferred stock / redemption value (Compustat Item #56) minus lag(preferr ed stock / redemption value). Standard errors are shown in parenthesis. Base choice is to maintain prior repurchase level. Variable Decrease Repurchase Increase Repurc hase Decrease Repurchase Increase Repurchase VIX 0.037 *** 0.001 0.044 *** 0.010 (0.01) (0.01) (0.02) (0.01) VIX + CFQuart2*VIX 0.027 0.003 0.045 ** 0.003 (0.02) (0.01) (0.02) (0.01) VIX + CFQuart3*VIX 0.021 0.000 0.034 ** 0.009 (0.02) (0.01) (0.02) (0.01) VIX + CFQuart4*VIX 0.010 0.022 *** 0.002 0.031 *** (0.02) (0.01) (0.02) (0.01) ReturnVolatility 0.006 ** 0.002 0.006 0.007 (0.00) (0.00) (0.01) (0.00) ReturnVolatility+ CFQuart2*ReturnVol 0.009 *** 0.001 0.007 ** 0.002 (0.00) (0.00) (0.00) (0.00) ReturnVolatility + CFQuart3*ReturnVol 0.010 *** 0.001 0.011 *** 0.005 (0.00) (0.00) (0.00) (0.00) ReturnVolatility + CFQuart4*ReturnVol 0.025 *** 0.013 *** 0.023 *** 0.015 *** (0.00) (0.00) (0.00) (0.00) AnalystDispersion 6.180 22.894 ** (7.80) (11.61) AnalystDispersion + CFQuart2*Disp 12.150 7.471 (9.69) (9.37) AnalystDispersion + CFQuart3*Disp 7.874 10.252 (8.32) (9.65) AnalystDispersion + CFQuart4*Disp 57.636 ** 25.356 (28.46) (26.96) ***,**,* denote significance at the 1%, 5%, and 10% levels, respectively

PAGE 74

74 Table 2 1 5 Logit regression for the c hoice to initiate a repurchase for p rior n onr epurchasers This table shows the logit regression and linear combination for interaction coefficients for the firms choice to either maintain or change their repurchase level given that they did not repurchase stock during the prior fisc al year. The firms repurchase level is defined as purchase of common and preferred stock (Compustat Item #115) plus the minimum of 0 and (preferred stock / redemption value (Compustat Item #56) minus lag(preferred stock / redemption value). Standard err ors are shown in parenthesis. Base choice is to maintain not repurchasing shares. Variable 1 2 NYP 0.409 0.271 (0.22) (0.22) MtoB 0.188 *** 0.247 *** (0.02) (0.03) dA/A 0.344 *** 0.598 *** (0.05) (0.09) E/A 0.026 0.005 (0.20) (0.23) VIX 0.004 0.007 (0.01) (0.02) ReturnVolatility 0.001 0.000 (0.00) (0.00) CashFlow/Assets 1.184 *** 2.049 *** (0.33) (0.50) LagNetDebt/Assets 1.010 *** 1.076 *** (0.08) (0.15) NegativeRetainedEarn 0.359 *** 0.203 *** (0.04) (0.07) ChangeDivTaxRate 0.032 0.238 (0.81) (0.92) RepatTaxCutDummy 0.222 ** 0.254 (0.09) (0.14) AnalystDispersion 2.723 (3.05) LagReturn 0.001 0.008 (0.01) (0.01) ***,**,* denote significance at the 1%, 5%, and 10% levels, respectively

PAGE 75

75 Table 2 15 Continued Variable 1 2 CFQuart2*VIX 0.012 *** 0.019 *** (0.00) (0.01) CFQuart3*VIX 0.026 *** 0.033 *** (0.00) (0.01) CFQuart4*VIX 0.032 *** 0.033 *** (0.01) (0.01) CFQuart2*ReturnVol 0.001 0.003 (0.00) (0.00) CFQuart3*ReturnVol 0.005 *** 0.009 *** (0.00) (0.00) CFQuart4*ReturnVol 0.004 0.005 ** (0.00) (0.00) CFQuart2*Disp 12.798 ** (5.90) CFQuart3*Disp 0.158 (5.29) CFQuart4*Disp 6.043 (7.97) Constant 1.936 *** 1.453 ** (0.42) (0.60) Industry Dummies Y Y Year Clustered S.E. Y Y Number of observations 39,877 15,124 Adjusted R2 0.055 0.053 ***,**,* denote significance at the 1%, 5%, and 10% levels, respectively

PAGE 76

76 Table 21 6 Linear combination for the choice to maintain or increase repurchase level for prior nonrepurchasers. This table shows the linear combination of coefficients for the interaction variables from Table 21 5 for the fi rms choice to either maintain or change their repurchase level given that they did not repurchase stock during the prior fiscal year. The firms repurchase level is defined as purchase of common and preferred stock (Compustat Item #115) plus the minimum of 0 and (preferred stock / redemption value (Compustat Item #56) minus lag(preferred stock / redemption value). Standard errors are shown in parenthesis. Base choice is to maintain not repurchasing shares. Variable 1 2 VIX 0.004 0.007 (0.01) (0.02) VIX + CFQuart2*VIX 0.016 0.012 (0.02) (0.02) VIX + CFQuart3*VIX 0.030 ** 0.026 (0.01) (0.02) VIX + CFQuart4*VIX 0.035 ** 0.026 (0.01) (0.02) ReturnVolatility 0.001 0.000 (0.00) (0.00) ReturnVolatility+ CFQuart2*ReturnVol 0.002 ** 0.002 (0.00) (0.00) ReturnVolatility + CFQuart3*ReturnVol 0.006 *** 0.009 *** (0.00) (0.00) ReturnVolatility + CFQuart4*ReturnVol 0.005 *** 0.005 ** (0.00) (0.00) AnalystDispersion 2.723 (3.05) AnalystDispersion + CFQuart2*Disp 15.521 *** (4.74) AnalystDispersion + CFQuart3*Disp 2.881 (3.95) AnalystDispersion + CFQuart4*Disp 8.766 (7.40) ***,**,* denote significance at the 1%, 5%, and 10% levels, respectively

PAGE 77

77 Figure 2 1. Percentage of f irms d ecreasing/ e liminating d ividends vs p ercentage of f irms i ncreasing d ividends by year. This figure compares the percentage of firms either decreasing or eliminating their dividends to the percentage of firms increasing their dividends. The cutoff for decreasing/increasing is a change in dividends per share (adjusted for splits) of 5% or 5%, respectively. The dashed line represents the percentage of firms increasing their dividends and the solid line represents the percentage of firms decreasing their dividends.

PAGE 78

78 CHAPTER 3 PRICE DISCOVERY AND RECENT TRENDS IN EXT ENDEDHOURS TRADING What happens to stock prices after the closing bell rings at the New York Stock Exchange? Do all investors simply pack up and quit thinking about the stock price until the opening bell rings at 9:30 am the next trading day? Does information about the underlying company lay stagnant for the next seventeen and a half hours? Of course not! Information that can affect the companys stock price continues to flow no matter the time of the day. In fact, as the ease of information flow has improved exponentially over the past couple of decades thanks to the internet as well as increased globalization, the amount of information released outside of normal stock market trading hours has increased as well. How does this new information get reflected into the companys stock price outside of normal trading hours? Some stocks are traded on multiple stock exchang es allowing investors to trade on one exchange in hours when the other(s) may be closed. However, even if a stock is not cross listed it is possible to trade outside of normal trading hours. Starting in 1999 extendedhours trading was made available to all investors. I examine the changes and trends that have occurred in extendedhours trading since 1999. I also consider the impact extendedhours trading has had on the price discovery process throughout the day. Barclay and Hendershott (2003) provide the most detailed look to date into the extendedhours trading environment and the price discovery that occurs outside of normal trading hours. However, their sample consists of only NASDAQ trades for the year 2000. I expand on their sample to include all firms in the Standard and Poors 1500 Composite Index for the years 1999 to 2009. By significantly increasing the

PAGE 79

79 sample and including a timeseries aspect I am able to demonstrate the t rends that have occurred in extendedhours trading over time as well as give a much more complete overview of the extendedhours trading environment. Chordia, Roll and Subrahmanyam (201 1 ) show that there has been a recent trend in trading resulting in a very high volume of smaller trades. Turnover has actually increased through their sample (1999 to 2008) despite a significant decrease in the average trade size. The difference is more than made up by the fact that the volume of trades has increased so si gnificantly. I investigate whether this same trend has transferred to extendedhours trading. While I find that the volume of trading occurring outside of regular trading hours has steadily increased at a rather strong pace, I do not find that the averag e trade size has behaved similarly to the behavior Chordia, Roll and Subrahmanyam (2011 ) find for all trading. However, beginning in late 2008 / early 2009 a pattern does begin to arise which suggests that extendedhours trading may be starting to mimic t he high volume, small trade size pattern. Similar to Barclay and Hendershott (2003) I examine the amount of price discovery that takes place before the market opens and after the market closes. Given my time series dataset I have the opportunity to also look at how this price discovery process changes over the years. I find that there has been a significant shift in price discovery throughout the day such that the nontrading hours are now nearly as important for the price discovery process as are the r egular trading hours. In fact, using Barclay and Hendershotts (2003) Weighted Price Contribution measure I demonstrate that the nontrading hours actually account for over half (57.52%) of the total price discovery for Standard & Poors 500 LargeCap Index firms (an increase from only

PAGE 80

80 10.11% in 1999). I utilize both Barclay and Hendershotts (2003) Weighted Price Contribution (WPC) measure and a newly created Absolute Price Discovery (APD) measure, which should be a better indicator of actual price discovery during time intervals, to investigate the shifts in price discovery over time. Even after controlling for a potential upward bias of the WPC, I still demonstrate a significant increase in price discovery outside of trading hours. Using my APD measur e the extendedhours price discovery increases from 5.71% in 1999 to 25.28% in 2009. The impact is even more profound for large market capitalization stocks that comprise the S&P 500 with an increase from 6.05% in 1999 to 41.09% in 2009. Both measures cl early demonstrate the increased importance of extendedhours trading in the formation of price changes throughout the day. The increased price discovery occurring prior to 9:30 am and after 4:00 pm can likely be attributed to many factors. One such potential factor is an increased propensity to announce earnings outside of trading hours. Given the amount of new information that is usually contained in an earnings announcement it follows that increased price discovery is going to occur immediately following these announcements. As firms become more likely to announce outside of trading hours the amount of price discovery occurring outside of trading hours is increased on these dates. Other potential factors that may be aiding the shift in price discovery include improved ease of information flow at all times of the day, rapidly growing importance of the global market, and easier access to extendedhours trading thr o u gh improved trading technology.

PAGE 81

81 Overview of the Existing Extended Hours Trading Literature The early literature on extendedhours trading mostly focused on two distinct areas. The first area deals w ith the changes in stock value that occurred during the nontrading period which was reflected at the opening of trade the next day. Examples of papers in this area include Oldfield and Rogalski (1980) and Houston and Ryngaert (1992). Oldfield and Rogalski (1980) demonstrate that stock returns follow a different jump process during overnights, holidays, weekends and holiday weekends. Houston and Ryngaert (1992) show that weekly volume and stock return variance remains relatively unchanged during weeks with reduced trading hours. In these cases they find that the volume and variance is instead shifted to the trading days following the closing of the stock market. The second area focuses on stocks that are listed on at least one stock market outside of the U.S. as well as being listed on one of the major U.S. exchanges. For example, Neumark, Tinsley, and Tosini (1991) found that price changes on the U.S. exchanges were adequately incorporated the next day in the international markets, but the opposite wasnt always the case. However, extendedhours trading has come a long way since these early articles T rading jointly listed stocks on an alter native exchange is no longer the only manner of trading a stock after hours and looking at the opening of trading the next trading day is not the only way to observe stock price changes outside of trading hours Extended hours trading began in 1975 but was originally limited to only institutional investors and large block trades. It wasnt until 1999 that regular investors were allowed to trade outside of the normal 9:30 am to 4 :00 pm trading day. Due to this fundamental change, I will focus on the more modern strand of the literature which looks at the extendedhours market since the rule change in 1999.

PAGE 82

82 One of the most important papers in the current extendedhours literature is Barclay and Hendershott (2003). This paper is not only important because o f its overall findings, but also because it has the most comprehensive summary statistics concerning extendedhours trading. The authors show a clear picture of extendedhours trading relative to regular trading hours. The data used by Barclay and Hendershott (2003) contains all after hours trades and quotes for N ASDAQ listed stocks for 212 trading days during 2000. They show that these stocks collectively average around 25,000 after hours trades per day. This represents almost 4% of the daily total trading volume on average. Throughout their study they focus on the 250 highest volume stocks from their sample, which they show represents about 75% of all after hours trading. Barclay and Hendershott (2003) demonstrate in Figure 1 of their paper that after hours trading is strongest directly before the open and after the close of regular trading hours. It also demonstrates that volatility follows the same pattern, but wit h a much less severe dropoff during after hours trading. The main finding of Barclay and Hendershott (2003) is that the amount of information on a per trade basis during extendedhours trading is significantly higher than during regular trading hours. T herefore, even though the volume of trading is significantly lower, there is still strong price discovery outside of regular trading hours. Barclay and Hendershott (2003) also make other interesting observations about extendedhours trading. First, they argue that the lack of popularity of extendedhours trading likely stems from their finding that there is a higher probability of informationbased trades on a per trade basis outside of regular hours. Smaller liquidity traders prefer to trade together to minimize the likelihood of trading against informed

PAGE 83

83 traders and are therefore better off pooling together during regular trading hours when the per trade likelihood of informed trading is lower and trading costs are smaller. They also touch on how the dif ferent characteristics of extendedhours trading could affect whether or not a firm releases earnings announcement s during regular trading hours or after hours. In their paper they claim, The noisier stock prices and less efficient price discovery after hours could affect firms decisions about the timing of their public announcements, such as earning announcements. Announcements made after hours are likely to generate greater volatility and larger price reversals than are announcements made during the t rading day. (Barclay and Hendershott, 2003, p. 1070) The idea of extendedhours trading affecting firms decisions on public announcements has generated some interest in recent years including Greene and Watts (1996), Bagnoli Clement and Watts (2006), and others. In Barclay and Hendershott (2004) the authors use essentially the same sample as Barclay and Hendershott (2003) to look at extendedhours trading and its effect on market microstructure characteristics. They look specifically at how the lack of trading outside of normal trading hours affects trading costs through bidask spreads. Their results are consistent with what should be expected. The lack of liquidity in the extendedhours market results in higher trading costs. Both quoted and eff ective spreads increase significantly outside of normal trading hours. The percentage effective half spread moves from an average of approximately 0.17% during the trading day to approximately 0.6% after hours. They also find that spreads become signific antly larger for stocks with lower trading volume. Barclay and Hendershott (2004) then break the bid ask spread down into its three fundamental components: inventory holding costs,

PAGE 84

84 order processing costs, and adverse selection costs.15 They find that the adverse selection component of the spread is 15 times larger during the preopen than during normal trading hours and 7 times larger during the post close than during normal trading hours. They sum up why they feel extendedhours trading will never grow m uch beyond its current low level of trading with the following thoughts: The magnitude of the liquidity externalities suggest that exchanges have little incentive to expand their trading hours due to competitive pressure. Despite the wide spreads, profit opportunities for dealers to provide liquidity appear limited and the high adverse selection and low trading activity make monitoring the market costly. The wide spreads should discourage investors from trading after hours unless they have very high liqui dity demands or short lived private information. Finally, the investor protections, for example, warnings of high trading costs and volatility, currently employed by brokers and regulators s hould be continued. (Barclay and Hendershott, 2004, p. 709) There are relatively few other papers that look at the current extendedhours trading market. One such paper that does is Zdorovtsov (2003). Zdorovtsov (2003) looks mostly at the volatility over the extendedhours period and considers the private and public i nformation hypotheses. One of the main findings of the paper is that a large amount of trading volume in the preopen period coincides with higher volatility in overnight returns and lower volatility during regular trading hours. According to the author this represents a shift in the price discovery toward the preopen hours. Another important finding of the paper is that the greater the flow of public information after hours the greater the after hours volatility (and the same is true for during normal trading hours). Finally, Zdorovstov (2003) finds evidence (as in prior studies) that information releases outside of trading hours are of greater economic significance than information releases during trading hours. 15 As found in Stoll (1989)

PAGE 85

85 Data The core data for the panel utiliz ed in this study consists of all trades made prior to 9:30 am or after 4:00 pm for the years 1999 thr o u gh 2009. The trade by trade data is from the N ew Y ork S tock E xchange Trades and Quote (TAQ) database. Along with the exact date, time, price and volume of the trade, the TAQ data also identifies the exchange on which the trade occurred as well as distinguishing the condition of the trade. By utilizing the condition code from the TAQ database, I eliminate all trades that may not represent an actual extendedhours trade or may not actually be contributing to the true price discovery process.16 The panel of extended hours trades is linked to the firms CUSIP using the TAQ M aster file from Wharton Research Data Services (WRDS). The data is then matched to the constituents of the Standard and Poors Composite 1500 index (S&P 1500) using Compustats Index Constituents database. Firms are only maintained in the sample if they were included in the S&P 1500 for at least one day of the year. By limiting the sample to only the S&P 1500 I keep the sample from being too large while still keeping a broad representation of the United States stock market and allowing comparison of larg e market capitalization (S&P 500), midsize market capitalization (S&P 400), and small market capitalization (S&P 600) stocks. The data is then matched to the Center for Research in Security Prices (CRSP) database. From CRSP I obtain the firms primary e xchange, monthly trade volume and shares outstanding. 16 For example, trades with a condition code of W are eliminated due to the fact they may improperly affect the appearances of the pricediscovery process. These trades are defined as A trade where the price repor ted is an average of the prices for transactions during all or any portion of the trading day. An example of trades that are eliminated due to fear that they are not actual extendedhours trades are trades with a condition code of Z. These trades are defined as A transaction that is reported to the tape at a time later than it occurred and when other trades occurred between the time of the transaction and its report time. (New York Stock Exchange, 2008)

PAGE 86

86 Trends in the ExtendedHours Trading Environment Chordia, Roll and Subrahmanyam (201 1 ) show that there has been a trend toward increased share turnover despite a dropoff in the average trade size. The cause for this, as demonstrated in Chordia, Roll and Subrahmanyam (201 1 ) is that stock trading has shifted towards a larger concentration of very frequent, but small trades. They document this tr end by looking at trades of NYSE stocks for a sample period from 1993 to 2008 using TAQ data. In a similar manner, I look to see whether a comparable pattern has occurred during the nontrading hours. Though my sample period is shorter due to the fact that regular investors were not given access to extendedhou rs trading until 1999, my panel data spans a significant enough portion of Chordia, Roll and Subramanyams (201 1 ) sample period that I should be able to document a similar pattern if it exists. Similar to Chordia, Roll and Subrahmanyam (201 1 ) I calculat e the monthly average turnover by firm. However, instead of using monthly trading volume as the numerator I use monthly extendedhours trading volume. This allows me to look at the monthly turnover that occurs outside of normal trading hours. Monthly ex tendedhours turnover for firm i during month j is calculated as the total number of shares traded outside of normal trading hours for firm i during month j divided by the total number of shares outstanding for firm i at the end of month j Table 3 1 show s the monthly turnover across time. The data is also separated out into largecap (S&P 500), midcap (S&P 400) and smallcap (S&P 600) firms. Figure 3 1 looks at the data graphically. From Figure 3 1 it seems that the exponential increase in turnover t hat Chordia, Roll and Subrahmanyam (2011 ) demonstrate has occurred during regular trading hours has not necessarily

PAGE 87

87 transferred over to the extendedhours trading period as turnover was relatively stagnant from 2002 to 2006. However, there is a significant jump in extendedhours turnover from early 2007 to the middle of 2009 with the most dramatic spikes occurring early 2007 and late 2008 / early 2009. Another interesting observation from the turnover data is that there is not a significant difference in turnover between small cap, mid cap and largecap stocks in the extended hours trading period. This is somewhat surprising as one might expect that largecap firms would be more likely to be paid attention to outside of trading hours, resulting in higher extendedhours turnover than for smallcap firms. To look more closely at whether or not a similar pattern is occurring during extendedhours trading as is occurring during regular trading hours I look at the number of trades per month (Table 3 2 and Figure 3 2) as well as the size of the average trade (Table 3 2 and Figure 3 3). From Figure 3 2 it is evident that there is a significant upward trend in the number of extendedhours trades taking place over time. Similar to what Chordia, Roll and Subrahmanyam (201 1 ) identify during regular trading hours, a pattern of increased frequency of smaller trades is evident in the extendedhours environment. However, this pattern does not show until the last year of my sample, 2009. In 2009 the average t rade size drops from about 3,000 shares and a total value of approximately $100,000 to about 1,000 shares and a total value of approximately $25,000. However, this shift is not nearly as prolonged or as significant as the shift documented by Chordia, Roll and Subrahmanyam (201 1 ). Therefore it appears that the trend towards a higher frequency of smaller trades that has occurred over recent years has not necessarily had the same impact on the extendedhours trading environment.

PAGE 88

88 As in Barclay and Hendershott (2003) I break the trading down into 30minute intervals to measure the amount of trading volume across time for the nonregular trading hours. The results are shown in Table 3 3 and Figure 3 4. As would be expected, the majority of extendedhours trades occur close to the opening of trading at 9:30 am and the close of trading at 4:00 pm. From prior studies, such as Jain and Joh (1988), it is known that a ushaped pattern exists between 9:30 am and 4:00 pm. Therefore if I were to fill in the gap between 9:30 am and 4:00 pm it would likely show a huge jump in trading right at 9:30 am which would come down to create the left side of the u until approximately 1:00 pm and then turn back up creating the right side of the u reaching its peak at 4:00 pm. Fro m there it would show a large drop back down to the 4:00 pm to 4:30 pm level and a continued decrease through the overnight period. Additionally, I explore the breakdown of extendedhours trades by trading venue. Table 3 4 and Figure 3 5 demonstrate the p ercentage of trades classified as occurring on each individual venue by year. Around the turn of the Twenty First Century extendedhours trading was dominated by the NASDAQ with over 98% of all trades. From 2002 to 2005 the National Stock Exchange (NSX) appears quite relevant with approximately 25% of all trades. However, Chung and Kim (2009) state that the exchange code associated with NSX may be reflecting trades that occurred on the Island Exchange. Island was acquired in 2005 by NASDAQ which would explain the shift in trading volume from NSX to NASDAQ in 2006. Another large player is the Archipelago Exchange (ARCA) which became relevant in 2003. ARCA was acquired by the New York Stock Exchange in 2005. In recent years the extended hours market has been dominated by the NASDAQ and NYSE (through ARCA) exchanges with NASDAQ

PAGE 89

89 having 53.15% of the volume while ARCA has 40.31%. The only other exchange threatening the two major players as of 2009 is the BATS Exchange, an electronic crossing network ( ECN ) Price Discovery in Extended Hours Trading Potentially more interesting than just the surface trends that are occurring in extendedhours trading is the amount of price discovery that takes place outside of the normal 9:30 am to 4:00 pm trading sessions. As the financial world becomes increasingly a global entity we may expect to see a greater percentage of price discovery happening outside of regular trading hours. While the United States markets are closed information is still flowing both inside the U nited States and outside the United States. For example, f irms have been announcing earnings during the preopen or post close periods with increasing frequency. The large amount of potentially significant new information that can be revealed during an earnings announcement should likely shift some of the price discovery process into the nontrading periods on days around these extendedhours information releases. Another factor that may have a positive impact on the price discovery in extendedhours tra ding is the increasing ease of information flow at all hours of the day. When considering this increased flow of information with an always growing impact of the global market one can easily understand the potential for an increased importance of extendedtrading hours on the price discovery process. Barclay and Hendershott (2003) have previously looked at the price discovery process that occurs outside of normal trading hours. However, their sample only allows them to look at the amount of price discovery taking place during one singular year, 2000. It also only looks at NASDAQ stocks and only considers the top 250 by volume.

PAGE 90

90 Therefore, they do not capture changes in the price discovery over time, instead only capturing the amount of price discovery one year after regular investors were introduced t o after hours trading. I start by using methodology similar to that used in Barclay and Hendershott (2003). I recreate their Weighted Price Contribution (WPC) measure for each day and each time period, i such that WPC is defined as: = | | | | (3 1) where is the logarithmic return for stock s during period i and is the close to close return for stock s.17 Periods are segments of the trading day consisting of post close (closing price of the prior trading day to 6:30 pm the prior trading day), overnight (6:30 pm the prior trading day to 8:00 am), preopen (8:00 am to the last trade prior to the open of trading) and regular trading hours (last trade prior to the open of trading to the close of trading). The results using Barclay and Hendershotts (2003) WPC measure are shown in Table 3 5 and Figure 3 6. For the entire sample (the S&P 1500), the percentage of price discovery that occurs outside of trading hours increases from 8.49% in 1999 to 32.35% in 2009 according the WPC measure. The impact is even stronger for the large capitalization stocks that make up the S&P 500 as the percentage increases from 10.11% to 57.52%. These firms are likely to experience more price discovery outside of trading hours as they generally receive more attention from investors resulting in more 17 The methodology for the Weighted Price Contri bution (WPC) is first utilized in Barclay and Warner (1993). Several papers have looked at the effectiveness of this measure, including van Bommel (2009) and Wang and Yang (2010).

PAGE 91

91 liquidity, have more information being released throughout the day by the media and are more likely to have an international presence. Table 3 5 also breaks the price discovery in extendedhours trading down into three more specific segments: post close (4:00 pm to 6:30 pm), overnight (6:30 pm to 8:00 am) and preopen (8:00 am to 9:30 am). While it is evident that each time interval has gained significance in the price discovery process, the overnight period stands out as having the most growth. Starting in 1999 and moving thr o u gh the early 2000s the overnight period had essentially zero effect on the stock price. However, in the late 2000s it becomes nearl y as important for price discovery as the preopen period. This is especially true for the large market capitalization stocks that comprise the S&P 500 where the overnight period actually passes the preopen period for price discovery in 200 8 One potenti ally concerning issue with using the WPC measure from Barclay and Hendershott (2003) is the impact price reversals can have on the WPC for a specific time interval For example, consider a situation where stock XYZ closes day 0 at a price of $100.00. Dur ing the extendedhours periods between the close on day 0 and the last trade prior to the 9:30 am open on day 1 the stock price increases to $101.50. However, during regular trading hours new information is revealed that pushes the price back to $101.00. Using the WPC methodology of Barclay and Hendershott (2003) the second variable in Equation 31, , will be greater than 1 for the extendedhours period and negative for the regular trading hours. However, what occurs during the 9:30 am to 4:00 pm trading session is not necessarily negative price discovery; it is simply price discovery in the opposite direction of what occurred in extendedhours

PAGE 92

92 trading. Therefore, I create a new measure, Absolute Price Discovery (APD), which measures price discovery such that price movement during any given interval is considered positive price discovery regardless of the direction relative to the full time period. APD is measured as: APD= , (3 2) where is t he logarithmic return for stock s during period i This measure should better capture the true price discovery process that occurs within each time interval as it treats both positive and negative price movements as positive price discovery. The results using APD instead of WPC are shown in Table 3 6 and Figure 3 7 When utilizing the APD measure the proportion of price discovery that occurs during extendedhours trading is reduced. This is due to the fact that there are often days where the price discovery during regular trading hours will turn in the opposite direction of the price discovery that has already taken place leading up to the opening trade. By counting this as positive price discovery instead of negative price discovery it less ens the percentage impact of what has occurred outside of trading hours. However, even after eliminating the upward bias on the amount of price discovery taking place during extendedhours trading that is caused by the WPC measure, the APD measure still d emonstrates a strong shift in the price discovery process towards the nontrading hours. For the entire sample, the S&P 1500, the percentage of price discovery taking place increases from 5.71% in 1999 to 25.28% in 2009. As with the WPC measure, the nontrading hours price discovery is even larger for the large market capitalization S&P 500 stocks. The A PD for the S&P 500 increases from 6.05% in 1999 to 41.09% in 2009. Looking at the specific time intervals, post close, overnight and preopen, the

PAGE 93

93 patt ern remains quite similar to that of the WPC measure. However, the overnight period does not pass the preopen period in amount of price discovery as it does for the WPC measure. From both the WPC measure and the A PD measure it is quite apparent that the trading which occurs before the market opens at 9:30 am and after the market closes at 4:00 pm has become a large part of the price discovery process. Conclusion Extendedhours trading ha s been a mostly ignored topic in the academic finance literature. Barclay and Hendershott (2003) look closely at the after hours trading environment and the amount of price discovery that takes place outside of normal trading hours. However, their dataset is comprised of only one year, 2000, and is not far removed from 1999 when regular investors were first granted access to extendedhours trading. In this study I use a large dataset of all extendedhours trades of stocks in the S&P 1500 Composite Index from 1999 to 2009 to demonstrate how extendedhours trading has changed over time. While turnover has increased, the pattern over time does not mimic the regular trading hours pattern identified by Chordi a, Roll and Subrahmanyam (2010). They show that turnover has quickly increased since the early 2000s due to a shift towards a high volume of smaller trades. The volume of extendedhours trades has grown rapidly as well ; however, the shift in trade size has only recently appeared within the last few months of 2009. This shift is much too short lived to make any assumption that a similar pattern is beginning to surface outside of trading hours. A much more interesting trend appears when considering the price discovery process occurring during extendedhours trading as first investigated by Barclay and Hendershott (2003) They demonstrate that there is relatively low price discovery

PAGE 94

94 occurring outside of trading hours in the year 2000, even when only considering the largest 250 NASDAQ stocks by volume. However, the panel dataset I have created in this study allows me to look at the shifts in price discovery that have occurred in the past decade since Barclay and Hendershotts (2003) sample. As the financial world moves closer to a 24hour trading day, the hours outside of 9:30 am to 4:00 pm on Monday Friday have become i ncreasingly important for the price discovery process. Information is released at all times of the trading day and what happens in the international stock markets has an increasingly large impact on stocks in the United States. Around the turn of the Twe nty First C entury the price discovery taking place outside of trading hours made up only approximately 10% of the total price discovery for a 24hour period. However, by the late 2000s this number has increased to over 30% and even higher for large marke t capitalization stocks. I show that this result holds even after adjusting for the potential upward bias on the price discovery in nontrading hours that comes from using Barclay and Hendershotts (2003) Weighted Price Contribution measure. To adjust for this bias I use a newly created measure, the Absolute Price Discovery (APD) which allows price movement s to count as positive price discovery for a time interval even if it is against the direction of the overall return for the 24hour period. When using the APD measure the amount of price discovery occurring in extendedhours trading still shows a significant increase from 5.71% in 1999 to 25.28% in 2009. For S&P 500 largecap stocks this increase is even greater from 6.05% in 1999 to 41.09% in 200 9. My results clearly demonstrate the increased importance of nontrading hours as the United States financial markets become increasingly more a world market with a

PAGE 95

95 24hour trading period. While extendedhours trading has been mostly ignored in financial academic literature, I believe that the results of this study indicate that it deserves more attention in the future. In extensions of this paper I plan to look at the impact crosslistings have on the 24hour price discovery process over time. I believ e this will shed even more light on the shift in price discovery away from the prior concentration that occurred during the 9:30 am to 4:00 pm, Monday Friday trading days.

PAGE 96

96 Table 3 1. Average m onthly e xtendedh ours t urnover by y ear. This table shows average monthly extendedhours turnover by year. Extendedhours turnover is calculated as total number of shares traded outside of trading hours for the month divided by the firms total number of shares outstanding. 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 S&P 1500 Composite 0.006% 0.012% 0.017% 0.022% 0.024% 0.023% 0.019% 0.022% 0.043% 0.040% 0.037% S&P 500 Large Cap 0.004% 0.007% 0.012% 0.018% 0.022% 0.017% 0.018% 0.024% 0.037% 0.038% 0.038% S&P 400 Mid Cap 0.007% 0.015% 0.017% 0.024% 0.024% 0.023% 0.019% 0.023% 0.039% 0.038% 0.036% S&P 600 Small Cap 0.010% 0.017% 0.024% 0.027% 0.027% 0.028% 0.021% 0.019% 0.052% 0.042% 0.037%

PAGE 97

97 Table 3 2. Number of e xtendedh ours t rades p er y ear and a verage e xtendedh ours t rade size. This table shows the volume of extendedhours trades occurring per year (in thousands) for the S&P 1500 as well as the average extendedhours trade size in both number of shares and doll ar value. 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 Number of Trades Per Year (In Thousands) 702 2377 2977 3871 4434 4901 4930 6000 7827 10518 18547 Average Trade Size (Number of Shares) 2,092 1,864 2,409 2,533 2,587 2,226 2,334 2,694 3,772 3,051 2,077 Average Trade Size (Dollar Value) $138,663 $108,695 $82,028 $71,358 $76,607 $73,681 $84,055 $103,882 $147,544 $91,119 $40,964

PAGE 98

98 Table 3 3. Average d aily n umber of e xtendedh ours t rades by 30m inute i nterval This table shows the average daily number of extendedhours trades occurring during each 30minute interval for the S&P 1500. 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 Pre 7:00am 2.10 20.30 30.03 72.33 255.68 7am 7:30am 0.01 0.00 18.34 58.65 146.16 320.77 790.64 7:30am 8am 0.08 0.05 0.05 0.00 54.99 154.27 373.66 656.60 1,558.29 8am 8:30am 122.88 599.43 810.15 1,001.86 824.00 884.42 1,089.74 1,826.12 2,236.82 3,681.26 10,664.75 8:30am 9am 210.07 875.41 974.08 1,071.86 1,051.03 1,025.67 1,439.06 2,276.88 2,581.98 4,650.12 13,448.76 9am 9:30am 794.58 2,248.33 2,469.03 2,529.26 4,477.01 5,273.94 5,260.04 5,694.75 6,223.01 8,195.46 19,666.42 4pm 4:30pm 1,544.69 3,620.98 4,813.94 6,746.69 6,794.94 7,287.24 7,262.68 7,966.37 11,206.25 13,638.02 13,727.33 4:30pm 5pm 63.27 826.06 1,140.21 1,639.20 1,886.37 2,402.90 2,063.37 2,884.14 3,810.38 3,536.49 4,740.63 5pm 5:30pm 26.73 556.42 886.68 1,085.56 1,305.96 1,482.96 1,040.33 1,370.98 1,969.30 1,904.56 2,790.44 5:30pm 6pm 12.90 397.41 549.57 813.15 806.00 688.28 644.71 813.84 1,097.09 1,432.85 1,679.48 6pm 6:30pm 10.32 305.68 359.57 467.90 405.33 366.24 605.02 661.69 688.89 805.92 1,439.79 6:30pm 7pm 0.00 1.20 0.22 3.64 20.16 14.98 45.10 94.24 379.35 562.66 1,040.94 Post 7:00pm 3.65 22.90 20.69 36.47 84.00 439.67 2,116.05 1,795.91

PAGE 99

99 Table 3 4. Percentage of e xtendedh ours t rades o ccurring by v enue. This table shows the percentage of extended hours trades occurring by venue for the S&P 1500. Venue 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 NASDAQ 99.80% 99.46% 98.78% 78.06% 63.46% 41.75% 38.17% 60.34% 66.19% 62.14% 53.15% ARCA 0.01% 0.10% 0.22% 0.69% 15.01% 26.10% 30.92% 37.21% 29.61% 32.77% 40.31% NSX 0.00% 0.00% 0.00% 20.51% 20.97% 29.43% 30.68% 2.28% 0.28% 0.35% 0.02% NYSE 0.02% 0.00% 0.01% 0.00% 0.00% 0.00% 0.00% 0.02% 3.83% 3.57% 0.84% AMEX 0.00% 0.00% 0.00% 0.00% 0.00% 0.01% 0.01% 0.00% 0.00% 0.00% 0.00% Boston 0.00% 0.08% 0.27% 0.36% 0.25% 2.51% 0.08% 0.05% 0.01% 0.00% 0.29% ISE 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.02% 0.91% 1.03% Chicago 0.14% 0.30% 0.53% 0.34% 0.26% 0.15% 0.13% 0.09% 0.05% 0.02% 0.01% CBOE 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.02% 0.01% Philadelphia 0.03% 0.05% 0.20% 0.04% 0.06% 0.05% 0.02% 0.00% 0.01% 0.00% 0.00% BATS 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.21% 4.34%

PAGE 100

100 Table 3 5. Weighted Price Contribution by y ear This table shows the Weighted Price Contribution for each time interval by year. The Weighted Price Contribution is calculated as: WPC= | ret| | ret| ret, ret where is the logarithmic return for stock s during period i and is th e close to close return for stock s. Days with zero price change are discarded. Time Interval 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 Whole Sample Post Close 0.36% 0.60% 1.00% 1.05% 1.32% 1.54% 1.51% 1.33% 2.24% 1.55% 1.36% Overnight 0.00% 0.01% 0.01% 0.49% 2.25% 1.57% 2.81% 5.45% 11.43% 12.29% 14.87% Pre Open 8.13% 12.00% 13.93% 14.99% 19.60% 20.77% 18.85% 20.88% 21.02% 16.89% 16.12% Trading Hours 91.51% 87.39% 85.06% 83.47% 76.83% 76.13% 76.83% 72.34% 65.31% 69.27% 67.65% S&P 500 LargeCap Post Close 0.23% 0.85% 1.49% 1.78% 1.66% 1.93% 1.87% 2.04% 2.48% 1.52% 1.93% Overnight 0.01% 0.02% 0.04% 1.25% 6.34% 4.54% 7.86% 14.04% 22.00% 27.70% 33.12% Pre Open 9.88% 13.38% 15.05% 15.44% 20.38% 21.26% 23.37% 30.04% 23.61% 23.11% 22.47% Trading Hours 89.89% 85.75% 83.42% 81.53% 71.62% 72.27% 66.90% 53.89% 51.91% 47.67% 42.48% S&P 400 Mid Cap Post Close 0.19% 0.40% 0.80% 0.53% 1.04% 1.29% 1.23% 0.91% 1.92% 1.30% 1.28% Overnight 0.00% 0.00% 0.00% 0.04% 0.51% 0.45% 1.07% 2.65% 8.41% 8.63% 10.22% Pre Open 6.21% 10.14% 13.26% 14.09% 17.07% 18.74% 17.22% 20.53% 19.10% 15.97% 16.08% Trading Hours 93.60% 89.46% 85.94% 85.33% 81.39% 79.53% 80.48% 75.91% 70.57% 74.10% 72.42% S&P 600 SmallCap Post Close 0.36% 0.60% 1.00% 1.05% 1.32% 1.54% 1.51% 1.33% 2.24% 1.55% 1.36% Overnight 0.00% 0.01% 0.01% 0.49% 2.25% 1.57% 2.81% 5.45% 11.43% 12.29% 14.87% Pre Open 8.13% 12.00% 13.93% 14.99% 19.60% 20.77% 18.85% 20.88% 21.02% 16.89% 16.12% Trading Hours 91.51% 87.39% 85.06% 83.47% 76.83% 76.13% 76.83% 72.34% 65.31% 69.27% 67.65%

PAGE 101

101 Table 3 6. Absolute Price Discovery by y ear This table shows the Absolute Price Discovery for each time interval by year. The Absolute Price Discovery is calculated as APD= ret, ret, where is the logarithmic return for stock s during period i Time Interval 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 Whole Sample Post Close 1.14% 1.96% 2.91% 2.96% 4.56% 5.10% 7.13% 9.07% 9.52% 5.38% 4.68% Overnight 0.00% 0.01% 0.01% 0.19% 1.03% 0.80% 1.45% 2.60% 6.40% 6.85% 7.80% Pre Open 4.57% 6.43% 7.04% 8.57% 11.31% 12.05% 11.96% 14.45% 15.31% 12.14% 12.80% Trading Hours 94.29% 91.61% 90.05% 88.27% 83.11% 82.05% 79.46% 73.88% 68.78% 75.64% 74.72% S&P 500 LargeCap Post Close 0.65% 2.89% 5.14% 5.44% 7.06% 6.25% 9.43% 11.23% 9.76% 4.55% 5.17% Overnight 0.01% 0.01% 0.02% 0.48% 2.66% 2.03% 3.66% 6.20% 11.49% 14.59% 16.25% Pre Open 5.39% 6.82% 7.27% 8.81% 11.94% 12.70% 15.80% 21.43% 18.89% 18.00% 19.66% Trading Hours 93.95% 90.28% 87.57% 85.27% 78.35% 79.02% 71.11% 61.14% 59.86% 62.86% 58.91% S&P 400 Mid Cap Post Close 0.82% 1.09% 1.14% 0.57% 2.61% 4.09% 6.93% 9.92% 9.28% 3.64% 4.01% Overnight 0.00% 0.00% 0.00% 0.02% 0.14% 0.15% 0.40% 0.89% 4.11% 4.10% 4.47% Pre Open 3.62% 5.81% 6.83% 8.11% 9.62% 10.31% 9.79% 12.37% 13.01% 10.24% 10.85% Trading Hours 95.56% 93.10% 92.03% 91.31% 87.63% 85.45% 82.87% 76.83% 73.61% 82.01% 80.67% S&P 600 SmallCap Post Close 2.03% 1.62% 1.90% 2.22% 3.46% 4.71% 5.07% 6.43% 9.43% 7.27% 4.72% Overnight 0.00% 0.00% 0.00% 0.02% 0.04% 0.07% 0.12% 0.46% 3.44% 2.19% 2.74% Pre Open 4.38% 6.47% 6.96% 8.69% 11.95% 12.74% 9.93% 9.42% 13.70% 8.48% 8.19% Trading Hours 93.59% 91.91% 91.14% 89.07% 84.54% 82.48% 84.89% 83.69% 73.43% 82.06% 84.34%

PAGE 102

102 Figure 3 1. Turnover by m onth for the S&P 500, S&P 400 and S&P 600. This figure shows the average monthly extendedhours turnover by stock for the S&P 500, S&P 400 and S&P 600 across time. Extendedhours turnover is calculated as total number of shares t raded outside of trading hours for the month divided by the firms total number of shares outstanding.

PAGE 103

103 Figure 3 2. Number of e xtendedh ours t rades by m onth. This chart shows the number of extendedhours trades per month across time.

PAGE 104

104 Figure 3 3. Size of extendedh ours t rades by m onth. This chart shows the size of extendedhours trades per month across time. The average size is calculated both as the average number of shares per trade and the average dollar value per trade.

PAGE 105

105 Figure 3 4. Num ber of e xtendedhours t rades by 30m inute i nterval. This chart shows the number of extendedhours trades per 30minute interval broken into three time segments. The dotted line represents the years 1999 through 2002, the dashed line represents the years 2003 through 2006, and the solid line represents the years 2007 through 2009.

PAGE 106

106 Figure 3 5 Percentage of e xtended h ours t rades by v enue. This chart shows the percentage of extendedhours trades by interval across time. 0.00% 10.00% 20.00% 30.00% 40.00% 50.00% 60.00% 70.00% 80.00% 90.00% 100.00% 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 NASDAQ ARCA NSX NYSE AMEX Boston ISE Chicago CBOE Philadelphia BATS

PAGE 107

107 A B C D Figure 3 6. Weighted Price Contribution by year. These charts demonstrate the relative Weighted Price Contribution that occurs during the post close, overnight, preopen and trading hours intervals by year. A) Firms in the S&P 1500 Co mposite Index. B) Firms in the S&P 500 LargeCap Index. C) Firms in the S&P 400 Mid Cap Index. D) Firms in the S&P600 Small Cap Index. 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 Weighted Price Contribution Post Close Overnight Pre Open Trading Hours 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 Weighted Price Contribution Post Close Overnight Pre Open Trading Hours 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 Weighted Price Contribution Post Close Overnight Pre Open Trading Hours 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 Weighted Price Contribution Post Close Overnight Pre Open Trading Hours

PAGE 108

108 A B C D Figure 3 7. Absolute Price Discovery by y ear. These charts demonstrate the relative Absolute Price Discovery that occurs during the post close, overnight, preopen and trading hours intervals by year. A) F irms in the S&P 1500 Composite Index. B ) F irms in the S&P 500 LargeCap Index. C ) F irms in the S&P 400 Mid Cap Index. D ) F irms in the S&P600 Small Cap Index. 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 Weighted Price Contribution Post Close Overnight Pre Open Trading Hours 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 Weighted Price Contribution Post Close Overnight Pre Open Trading Hours 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 Weighted Price Contribution Post Close Overnight Pre Open Trading Hours 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 Weighted Price Contribution Post Close Overnight Pre Open Trading Hours

PAGE 109

109 CHAPTER 4 CONCLUSION AND FUTURE WORK In this study I focus on two separate topics in the finance literature: payout policy and extendedhours trading. The first part of the study, Chapter 2, looks at the impact of market level uncertainty on a firms payout policy decision. To proxy for uncertainty I uti lize the Chicago Board Options Exchange Volatility Index (VIX). I demonstrate that, even after controlling for a wide range of firm level determinants of payout policy from prior literature, the VIX has a significant impact on the payout policy decision f irms make. This impact differs based on the firms relative cash flow level and the type of payout (dividend or repurchase) Firms with low levels of cash flow are forced to take a more conservative approach to dividend payouts. Despite the fact that fi rms are very reluctant to decrease dividends due to the negative signal associated with a decrease, low cash flow firms become significantly more likely to decrease dividend levels during high VIX periods.1 However, firms with relatively high levels of cash flow are better able to sustain current dividend levels throughout the period of volatility. On the other hand, a different pattern appears when looking at the repurchase choice. Repurchases are not viewed as sticky as dividends are. Instead repurchase decisions are looked at as more of a onetime way to return cash to stockholders. Therefore I find that firms with high levels of cash flow actually utilize repurchases as a way to opportunistically take advantage of high volatility periods. I show t hat these high cash flow firms have an increased propensity to initiate a dividend repurchase when the 1 The dividend stickiness story of Lintner (1956) demonstrates that firms try to avoid dividend decreases as this is associated with a stock price penalty given investors interpret the decrease as a negative signal for future earnings.

PAGE 110

110 VIX is high. It appears that they are using the high volatility in stock prices to try and time repurchases when their stock price is low. In the second part of the study, Chapter 3, I turn the focus to extendedhours trading. Despite being a mostly overlooked area in the academic finance literature, I show that the nontrading hours have become an increasingly important part of the price discovery process for stocks. Since non institutional investors were first given access to trading outside of the normal 9:30 am to 4:00 pm trading day the percentage of price discovery taking place in these nonregular trading hours has grown significantly. I demonstr ate this using both a previously established measure of price discovery, the Weighted Price Contribution (WPC), and my own measure of price discovery, the Absolute Price Discovery (APD) Using the WPC measure utilized in Barclay and Hendershott (2003) I f ind that the percentage of price discovery occurring in extendedhours trading for firms in the S&P 1500 has increased from 8.49% in 1999 to 32.35% in 2009. However, the WPC may be upwardly biased if you consider any price movement during a time interval to be positive price discovery. Therefore, I also utilize my newly created APD measure to attempt to eliminate this bias. Using the APD measure I still find a significant increase in price discovery from 5.71% in 1999 to 25.28% in 2009. The increase in price discovery is even more significant for the largecapitalization stocks in the S&P 500 as it changes from 10.11% in 1999 to 57.52% in 2009 utilizing the WPC (6.05% to 41.09% using the APD). I also look at trends in the volume and size of extendedho urs trades that have occurred since 1999. Chordia, Roll and Subrahmanyam (201 1 ) show that there has been a shift towards a higher volume of smaller trades in regular trading hours during

PAGE 111

111 their sample period (1993 to 2008) While I find that there has been an increase in the volume of trades occurring outside of trading hours, I do not necessarily find the size of the average trade has decreased as they demonstrate. Therefore, it does not seem as though the same trends have necessarily transferred to extendedhour trading. In extensions of this study I plan to further investigate the causes for the shift in price discovery. By determining cross listings of the firms in my sample and linking each firm to the hours of the day which the firm is traded on any exchange I will be able to better identify the impact of cross listings on extendedhours price discovery. I also plan to attempt to identify the effect the recent financial crisis may have had on the impact of international stock movements on the movements of stocks trading at the U.S. exchanges. I believe that extendedhours trading is an area of the financial literature which still has a lot to be rev ealed. I plan to continue work in this area and try to help further develop the literature on extendedhours trading.

PAGE 112

112 APPENDIX GARCH (1,1) ESTIMATE OF THE VIX To allow the extension of my dataset to include pre1990 data I construct a GARCH (1,1) estima te of the VIX dating back to 1962. Prior literature, such as Engle (2001) and Hao and Zhang (2010), show that a GARCH (1,1) estimate of volatility for the S&P 500 is a good proxy for the VIX. Utilizing S&P 500 daily logarithmic returns I use GARCH (1,1) methodology to calculate a fitted model of: = 0 0006196 + (A 1) = 0 0763073 + 0 1976602 (A 2) where is the actual logarithmic return for day t is the error in the estimate for day t is the GARCH estimate of volatility for day t is the actual error from the estimate for day t 1 and is the actual GARCH estimate of volatility from day t 1 F or each day, the GARCH estimate is calculated using the actual estimate and error from the prior days GARCH estimate. For the first day of the sample is assumed to be 0 and is the actual variance over the entire sample period. The GARCH measure calculated using this methodology results in an estimate that has a correlation with the VIX of approximately 0 .913. This correlation is in line with prior research, such as Hao and Zhang (2010) who find a correlation of approximately 0 .92. Results from tests run on the GARCH sample are shown in Table A 1 through Table A 12

PAGE 113

113 Table A 1 Logit regression for the choice to pay or not pay a dividend. This table shows the logit regression coefficients for the firms choice to either pay or not pay a dividend. A firm is considered to have a positive dividend if the ex date dividend per share (Compustat Item #26) is positive. Standard errors are shown in parenthesis. Base choice is to not pay a dividend. Variable Whole Sample 1962 1989 1990 2009 NYP 2.689 *** 3.635 *** 2.297 *** (0.18) (0.13) (0.20) MtoB 0.609 *** 0.701 *** 0.287 *** (0.06) (0.06) (0.03) dA/A 0.694 *** 0.647 *** 0.938 *** (0.06) (0.08) (0.07) E/A 2.156 *** 1.867 *** 0.370 (0.58) (0.68) (0.29) GARCH 0.046 *** 0.042 ** 0.031 *** (0.01) (0.02) (0.01) ReturnVolatility 0.041 *** 0.046 *** 0.042 *** (0.00) (0.00) (0.00) CashFlow/Assets 4.825 *** 5.185 *** 3.913 *** (0.43) (0.52) (0.47) LagNetDebt/Assets 0.420 *** 1.123 *** 0.091 0.099 0.183 0.110 NegativeRetainedEarn 1.522 *** 1.536 *** 1.366 *** (0.08) (0.13) (0.04) ChangeDivTaxRate 0.089 *** 0.018 0.087 (0.02) (0.02) (0.62) RepatTaxCutDummy 0.802 *** 0.234 ** (0.14) (0.11) LagReturn 0.019 ** 0.008 0.054 *** (0.01) (0.01) (0.01) ***,**,* denote significance at the 1%, 5%, and 10% levels, respectively

PAGE 114

114 Table A 1 Continued Variable Whole Sample 1962 1989 1990 2009 CFQuart2*GARCH 0.013 0.020 0.023 (0.01) (0.01) (0.01) CFQuart3*GARCH 0.005 0.012 0.021 (0.01) (0.01) (0.01) CFQuart4*GARCH 0.003 0.002 0.023 (0.01) (0.01) (0.01) CFQuart2*ReturnVol 0.004 0.003 0.005 (0.00) (0.00) (0.00) CFQuart3*ReturnVol 0.000 0.003 0.006 ** (0.00) (0.00) (0.00) CFQuart4*ReturnVol 0.001 0.003 0.008 ** (0.00) (0.00) (0.00) Constant 2.650 *** 2.371 *** 2.296 *** (0.28) (0.37) (0.33) Industry Dummies Y Y Y Year Clustered S.E. Y Y Y Number of observations 118,968 56,441 62,404 Adjusted R2 0.412 0.423 0.394 ***,**,* denote significance at the 1%, 5%, and 10% levels, respectively

PAGE 115

115 Table A 2 Linear combination for the choice to pay or not pay a dividend. This table shows the linear combination of coefficients for the interaction variables from Table A 1 for the firms choice to either pay or not pay a dividend. A firm is considered to have a positive dividend if the ex date dividend per share (Compustat Item #26) is positive. Standard errors are shown in parenthesis. Base choice is to not pay a dividend. Variable Whole Sample 1962 1989 1990 2009 GARCH 0.046 *** 0.042 ** 0.031 *** (0.01) (0.02) (0.01) GARCH + CFQuart2*GARCH 0.034 ** 0.022 0.008 (0.02) (0.02) (0.01) GARCH + CFQuart3*GARCH 0.042 *** 0.030 0.011 (0.02) (0.02) (0.01) GARCH + CFQuart4*GARCH 0.044 ** 0.040 ** 0.009 (0.02) (0.02) (0.01) ReturnVolatility 0.041 *** 0.046 *** 0.042 *** (0.00) (0.00) (0.00) ReturnVolatility+ CFQuart2*ReturnVol 0.045 *** 0.049 *** 0.047 *** (0.00) (0.00) (0.00) ReturnVolatility + CFQuart3*ReturnVol 0.041 *** 0.042 *** 0.048 *** (0.00) (0.00) (0.00) ReturnVolatility + CFQuart4*ReturnVol 0.040 *** 0.043 *** 0.050 *** 0.004 0.004 0.004 ***,**,* denote significance at the 1%, 5%, and 10% levels, respectively

PAGE 116

116 Table A 3 Multinomial logit regression for the choice to maintain or change dividend for prior dividend payers. This table shows the multinomial logit regression for the firms choice to either maintain or change their dividend given that they paid a dividend during the prior fiscal year. The firms dividend is measured using the ex date dividends per share for the fiscal year (Compustat Item #26). Standard errors are shown in parenthesis. Base choice is to maintain dividend level. Variable Decrease Dividend Increase Dividend Decr ease Dividend Increase Dividend Decrease Dividend Increase Dividend Whole Sample 1962 1989 1990 2009 NYP 1.295 *** 0.558 *** 1.474 *** 0.567 *** 1.207 *** 0.611 *** (0.16) (0.09) (0.10) (0.11) (0.23) (0.09) MtoB 0.146 ** 0.180 *** 0.011 0.302 *** 0.290 *** 0.136 *** (0.06) (0.07) (0.09) (0.11) (0.07) (0.05) dA/A 1.160 *** 1.251 *** 1.570 *** 1.567 *** 0.455 ** 0.725 *** (0.15) (0.11) (0.20) (0.15) (0.18) (0.11) E/A 2.286 *** 10.257 *** 3.928 *** 14.262 *** 1.800 *** 4.257 *** (0.51) (1.14) (0.81) (1.60) (0.43) (1.02) GARCH 0.029 *** 0.033 0.041 ** 0.009 0.035 *** 0.054 ** (0.01) (0.02) (0.02) (0.03) (0.01) (0.02) ReturnVolatility 0.028 *** 0.013 *** 0.018 *** 0.012 ** 0.036 *** 0.013 ** (0.00) (0.00) (0.00) (0.01) (0.01) (0.01) CashFlow/Assets 4.381 *** 2.568 *** 4.975 *** 2.179 *** 1.791 ** 2.502 *** (0.83) (0.62) (1.05) (0.74) (0.85) (0.88) LagNetDebt/Assets 0.784 *** 0.474 *** 1.094 *** 0.195 0.421 0.817 *** (0.15) (0.08) (0.17) (0.10) (0.25) (0.10) NegativeRetainedEarn 0.526 *** 0.431 *** 0.587 *** 0.251 0.598 *** 0.349 *** (0.07) (0.10) (0.13) (0.15) (0.07) (0.11) ChangeDivTaxRate 0.001 0.005 0.000 0.018 2.682 *** 0.044 (0.01) (0.01) (0.01) (0.02) (0.39) (0.42) RepatTaxCutDummy 0.395 ** 0.319 ** 0.230 0.237 ** (0.19) (0.13) (0.15) (0.11) ***,**,* denote significance at the 1%, 5%, and 10% levels, respectively

PAGE 117

117 Table A 3 Continued Variable Decrease Dividend Increase Dividend Decrease Dividend Increase Dividend Decrease Dividend Increase Dividend Whole Sample 1962 1989 1990 2009 LagReturn 0.095 *** 0.034 ** 0.022 0.018 0.631 *** 0.238 *** (0.03) (0.02) (0.01) (0.01) (0.07) (0.04) CFQuart2*GARCH 0.033 *** 0.004 0.060 *** 0.008 0.024 0.024 (0.01) (0.01) (0.01) (0.01) (0.01) (0.02) CFQuart3*GARCH 0.055 *** 0.017 0.074 *** 0.021 0.049 ** 0.038 ** (0.02) (0.01) (0.02) (0.01) (0.02) (0.02) CFQuart4*GARCH 0.018 0.030 ** 0.011 0.048 ** 0.040 ** 0.031 ** (0.01) (0.01) (0.03) (0.02) (0.02) (0.02) CFQuart2*ReturnVol 0.002 0.010 ** 0.007 0.000 0.005 0.024 *** (0.00) (0.00) (0.00) (0.00) (0.01) (0.01) CFQuart3*ReturnVol 0.002 0.008 ** 0.008 0.003 0.008 0.023 *** (0.00) (0.00) (0.01) (0.00) (0.01) (0.01) CFQuart4*ReturnVol 0.001 0.006 0.006 0.007 0.007 0.014 ** (0.00) (0.00) (0.01) (0.01) (0.01) (0.01) Constant 0.726 ** 0.847 *** 0.335 1.379 *** 2.159 *** 0.528 (0.36) (0.27) (0.37) (0.45) (0.45) (0.34) I ndustry Dummies Y Y Y Y Y Y Year Clustered S.E. Y Y Y Y Y Y Number of observations 52,347 52,347 33,592 33,592 18,755 18,755 A djusted R2 0.161 0.161 0.179 0.179 0.157 0.157 ***,**,* denote significance at the 1%, 5%, and 10% levels, respectively

PAGE 118

118 Table A 4 Linear combination for the choice to maintain or change dividend for pr ior dividend payers. This table shows the linear combination of coefficients for the interaction variables from Table A 3 for the firms choice to either maintain or change their dividend given that they paid a dividend during the prior fiscal year. The firms dividend is measured using the ex date dividends per share for the fiscal year (Compustat Item #26). Standard errors are shown in parenthesis. Base choice is to maintain dividend level. Variable Decrease Dividend Increase Dividend Decrease Dividend Increase Dividend Decrease Dividend Increase Dividend Whole Sample 1962 1989 1990 2009 GARCH 0.029 *** 0.033 0.041 ** 0.009 0.035 *** 0.054 ** (0.01) (0.02) (0.02) (0.03) (0.01) (0.02) GARCH + CFQuart2*VIX 0.004 0.028 0.019 0.017 0.011 0.030 (0.01) (0.02) (0.01) (0.03) (0.01) (0.02) GARCH + CFQuart3*VIX 0.027 0.016 0.032 ** 0.012 0.014 0.016 (0.02) (0.02) (0.02) (0.03) (0.02) (0.01) GARCH + CFQuart4*VIX 0.011 0.003 0.031 0.039 0.005 0.023 (0.01) (0.02) (0.02) (0.03) (0.01) (0.01) ReturnVolatility 0.028 *** 0.013 *** 0.018 *** 0.012 ** 0.036 *** 0.013 ** (0.00) (0.00) (0.00) (0.01) (0.01) (0.01) ReturnVolatility+ CFQuart2*ReturnVol 0.027 *** 0.003 0.025 *** 0.012 *** 0.031 *** 0.011 *** (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) ReturnVolatility + CFQuart3*ReturnVol 0.026 *** 0.005 0.026 *** 0.008 ** 0.028 *** 0.010 *** (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) ReturnVolatility + CFQuart4*ReturnVol 0.028 *** 0.008 *** 0.024 *** 0.005 0.029 *** 0.001 (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) ***,**,* denote significance at the 1%, 5%, and 10% levels, respectively

PAGE 119

119 Table A 5 Logit regression for the choice to maintain or pay a dividend for prior dividend nonpayers. This table shows the logit regression for the firm s choice to either maintain or change their dividend given that they did not pay a dividend during the prior fiscal year. The firms dividend is measured using the ex date dividends per share for the fiscal year (Compustat Item #26). Standard errors are shown in parenthesis. Base choice is to maintain no dividend. Variable Whole Sample 1962 1989 1990 2009 NYP 1.003 *** 1.462 *** 1.054 *** (0.14) (0.18) (0.16) MtoB 0.594 *** 0.897 *** 0.234 *** (0.14) (0.13) (0.06) dA/A 0.571 *** 0.319 1.189 *** (0.16) (0.13) (0.33) E/A 4.827 *** 3.974 *** 4.211 *** (0.71) (1.05) (0.94) GARCH 0.056 ** 0.037 0.069 *** (0.02) (0.03) (0.02) ReturnVolatility 0.010 *** 0.015 *** 0.007 (0.00) (0.00) (0.00) CashFlow/Assets 1.917 ** 3.448 *** 0.916 (0.95) (0.86) (1.24) LagNetDebt/Assets 0.969 *** 1.182 *** 1.022 *** (0.11) (0.13) (0.19) NegativeRetainedEarn 0.807 *** 0.988 *** 0.543 *** (0.13) (0.19) (0.15) ChangeDivTaxRate 0.030 0.045 5.176 *** (0.04) (0.03) (0.58) RepatTaxCutDummy 0.079 0.233 (0.29) (0.30) LagReturn 0.008 0.001 0.028 ** (0.01) (0.01) (0.01) ***,**,* denote significance at the 1%, 5%, and 10% levels, respectively

PAGE 120

120 Table A 5 Continued Variable Whole Sample 1962 1989 1990 2009 CFQuart2*GARCH 0.004 0.008 0.015 (0.01) (0.02) (0.02) CFQuart3*GARCH 0.005 0.001 0.018 (0.01) (0.01) (0.02) CFQuart4*GARCH 0.016 0.018 0.008 (0.01) (0.02) (0.02) CFQuart2*ReturnVol 0.001 0.008 0.006 (0.00) (0.00) (0.01) CFQuart3*ReturnVol 0.006 0.013 *** 0.002 (0.00) (0.00) (0.00) CFQuart4*ReturnVol 0.008 ** 0.009 ** 0.003 (0.00) (0.00) (0.01) Constant 1.034 ** 1.253 1.183 (0.45) (0.67) (0.54) Industry Dummies Y Y Y Year Clustered S.E. Y Y Y Number of observations 66,595 22,842 43,608 Adjusted R2 0.131 0.153 0.116 ***,**,* denote significance at the 1%, 5%, and 10% levels, respectively

PAGE 121

121 Table A 6 Linear combination for the choice to maintain or pay a dividend for prior dividend nonpayers. This table shows the linear combination of coefficients for the interaction variables from Table A 5 for the firms choice to either m aintain or change their dividend given that they did not pay a dividend during the prior fiscal year. The firms dividend is measured using the ex date dividends per share for the fiscal year (Compustat Item #26). Standard errors are shown in parenthesis Base choice is to maintain no dividend. Variable Whole Sample 1962 1989 1990 2009 GARCH 0.056 ** 0.037 0.069 *** (0.02) (0.03) (0.02) GARCH + CFQuart2*GARCH 0.051 ** 0.045 0.054 *** (0.02) (0.03) (0.02) GARCH + CFQuart3*GARCH 0.050 ** 0.036 0.051 *** (0.02) (0.03) (0.01) GARCH + CFQuart4*GARCH 0.040 0.018 0.061 *** (0.02) (0.02) (0.01) ReturnVolatility 0.010 *** 0.015 *** 0.007 (0.00) (0.00) (0.00) ReturnVolatility+ CFQuart2*ReturnVol 0.009 *** 0.007 0.013 ** (0.00) (0.00) (0.01) ReturnVolatility + CFQuart3*ReturnVol 0.004 0.002 0.009 *** (0.00) (0.00) (0.00) ReturnVolatility + CFQuart4*ReturnVol 0.002 0.006 ** 0.004 (0.00) (0.00) (0.00) ***,**,* denote significance at the 1%, 5%, and 10% levels, respectively

PAGE 122

122 Table A 7 Logit regression for the choice to repurchase shares or not repurchase shares. This table shows the logit regression and linear combination for interaction coefficients for the firms choice to either repurchase shares or not repurchase shares. A firm is considered to be a repurchaser if total repurchases are positive, where total purchases are defined as purchase of common and preferred stock (Compustat Item #115) plus the minimum of 0 and (preferred stock / redemption value (Compustat Item #56) minus lag(preferred stock / redemption value). Standard errors are shown in parenthesis. Base choice is to not repurchase shares. Variable Whole Sample 1962 1989 1990 2009 NYP 0.385 ** 0.094 0.852 *** (0.16) (0.16) (0.21) MtoB 0.146 *** 0.346 *** 0.148 *** (0.02) (0.06) (0.02) dA/A 0.590 *** 0.446 *** 0.613 *** (0.04) (0.06) (0.05) E/A 0.610 *** 1.767 *** 0.074 (0.23) (0.55) (0.20) GARCH 0.024 ** 0.033 *** 0.003 (0.01) (0.01) (0.01) ReturnVolatility 0.004 *** 0.005 *** 0.003 ** (0.00) (0.00) (0.00) CashFlow/Assets 1.228 *** 0.159 2.136 *** (0.28) (0.42) (0.30) LagNetDebt/Assets 0.859 *** 0.791 *** 1.051 *** (0.05) (0.06) (0.07) NegativeRetainedEarn 0.567 *** 0.625 *** 0.499 *** (0.04) (0.09) (0.03) ChangeDivTaxRate 0.242 ** 0.207 ** 0.331 (0.10) (0.09) (0.78) RepatTaxCutDummy 0.526 *** 0.311 ** (0.12) (0.12) PreSafeHarbor 0.914 *** 0.698 *** (0.13) (0.14) LagReturn 0.027 *** 0.027 *** 0.026 *** (0.00) (0.01) (0.01) ***,**,* denote significance at the 1%, 5%, and 10% levels, respectively

PAGE 123

123 Table A 7 Continued Variable Whole Sample 1962 1989 1990 2009 CFQuart2*GARCH 0.023 *** 0.025 *** 0.023 *** (0.00) (0.00) (0.01) CFQuart3*GARCH 0.038 *** 0.018 *** 0.048 *** (0.00) (0.01) (0.01) CFQuart4*GARCH 0.045 *** 0.034 *** 0.066 *** (0.01) (0.01) (0.01) CFQuart2*ReturnVol 0.004 *** 0.002 0.005 *** (0.00) (0.00) (0.00) CFQuart3*ReturnVol 0.007 *** 0.000 0.012 *** (0.00) (0.00) (0.00) CFQuart4*ReturnVol 0.008 *** 0.003 ** 0.014 *** (0.00) (0.00) (0.00) Constant 0.902 *** 0.433 0.988 *** (0.21) (0.21) (0.32) Industry Dummies Y Y Y Year Clustered S.E. Y Y Y Number of observations 118,992 56,445 62,547 Adjusted R2 0.101 0.067 0.131 ***,**,* denote significance at the 1%, 5%, and 10% levels, respectively

PAGE 124

124 Table A 8 Linear combination for the choice to repurchase shares or not repurchase shares. This table shows the linear combination of coefficients for the interaction variables from Table A 7 for the firms choice to either repurchase shares or not repurchase shares. A firm is considered to be a repurchaser if total repurchases are positive, where total purchases are defined as purchase of common and preferred stock (Compustat Item #115) plus the minimum of 0 and (preferred stock / redemption value (Compustat Item #56) minus lag(preferred stock / redemption value). Standard errors are shown in parenthesis. Base choice is to not repurchase shares. Variable Whole Sample 1962 1989 1990 2009 GARCH 0.024 ** 0.033 *** 0.003 (0.01) (0.01) (0.01) GARCH + CFQuart2*GARCH 0.046 *** 0.058 *** 0.026 (0.01) (0.01) (0.02) GARCH + CFQuart3*GARCH 0.062 *** 0.051 *** 0.051 *** (0.01) (0.01) (0.02) GARCH + CFQuart4*GARCH 0.069 *** 0.068 *** 0.068 *** (0.01) (0.01) (0.02) ReturnVolatility 0.004 *** 0.005 *** 0.003 ** (0.00) (0.00) (0.00) ReturnVolatility+ CFQuart2*ReturnVol 0.008 *** 0.007 *** 0.008 *** (0.00) (0.00) (0.00) ReturnVolatility + CFQuart3*ReturnVol 0.011 *** 0.005 *** 0.015 *** (0.00) (0.00) (0.00) ReturnVolatility + CFQuart4*ReturnVol 0.013 *** 0.008 *** 0.017 *** (0.00) (0.00) (0.00) ***,**,* denote significance at the 1%, 5%, and 10% levels, respectively

PAGE 125

125 Table A 9 Multinomial logit regression for the choice to maintain or change repurchase level for prior repurchasers. This table shows the multinomial logit regression and linear combination for interaction coefficients for the firms choice to either maintain or change their repurchase level given that they repurchased stock during the prior fiscal year. The firms repurchase level is defined as purchase of common and preferred stock (Compustat Item #115) plus the minimum of 0 and (preferred stock / redemption value (Compustat Item #56) minus lag(preferred stock / redemption value). Standard errors are shown in parenthesis. Base choice is to maintain repurchase level. Variable Decrease Repurchase Increase Repurchase Decrease Repurchase Increase Repurchase Decrease Repurchase Increase Repurchase Whole Sample 1962 1989 1990 2009 NYP 0.420 *** 0.114 0.047 0.274 ** 0.642 *** 0.298 ** (0.14) (0.10) (0.14) (0.12) (0.18) (0.12) MtoB 0.040 0.074 ** 0.115 0.059 0.003 0.044 (0.03) (0.04) (0.09) (0.08) (0.03) (0.04) dA/A 0.736 *** 0.003 0.746 *** 0.134 0.699 *** 0.054 (0.17) (0.17) (0.23) (0.25) (0.22) (0.22) E/A 1.421 *** 0.425 2.124 *** 0.401 1.074 *** 0.621 (0.34) (0.46) (0.69) (1.06) (0.41) (0.49) GARCH 0.020 0.008 0.022 0.003 0.033 *** 0.002 (0.01) (0.01) (0.02) (0.02) (0.01) (0.01) ReturnVolatility 0.008 *** 0.003 0.006 0.000 0.007 ** 0.001 (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) CashFlow/Assets 1.619 *** 0.104 0.906 0.007 2.132 *** 0.208 (0.47) (0.50) (1.00) (1.18) (0.49) (0.58) LagNetDebt/Assets 0.580 *** 0.297 *** 0.721 *** 0.156 0.654 *** 0.292 ** (0.08) (0.10) (0.16) (0.13) (0.10) (0.14) NegativeRetainedEarn 0.082 0.062 0.025 0.150 0.055 0.077 (0.09) (0.07) (0.19) (0.20) (0.10) (0.08) ChangeDivTaxRate 0.482 0.567 0.162 2.127 *** 1.478 0.422 ** (0.69) (0.73) (0.39) (0.71) (0.92) (0.18) RepatTaxCutDummy 0.529 *** 0.063 0.381 *** 0.001 (0.14) (0.10) (0.14) (0.10) ***,**,* denote significance at the 1%, 5%, and 10% levels, respectively

PAGE 126

126 Table A 9 Continued Variable Decrease Repurchase Increase Repurchase Decrease Repurchase Increase Repurchase Decrease Repurchase Increase Repurchase Whole Sample 1962 1989 1990 2009 PreS afeHarbor 0.231 *** 0.060 0.001 0.174 (0.09) (0.08) (0.07) (0.09) LagReturn 0.007 0.059 *** 0.033 0.050 0.016 0.064 ** (0.02) (0.02) (0.03) (0.04) (0.04) (0.03) CFQuart2*GARCH 0.001 0.006 0.029 0.019 0.006 0.007 (0.01) (0.01) (0.02) (0.02) (0.01) (0.01) CFQuart3*GARCH 0.022 ** 0.013 0.040 ** 0.040 0.012 0.004 (0.01) (0.01) (0.02) (0.03) (0.01) (0.01) CFQuart4*GARCH 0.042 *** 0.025 *** 0.039 0.025 0.048 *** 0.031 *** (0.01) (0.01) (0.02) (0.03) (0.01) (0.01) CFQuart2*ReturnVol 0.001 0.004 0.004 0.005 0.003 0.001 (0.00) (0.00) (0.01) (0.01) (0.00) (0.00) CFQuart3*ReturnVol 0.008 *** 0.006 ** 0.013 *** 0.016 ** 0.003 0.000 (0.00) (0.00) (0.00) (0.01) (0.00) (0.00) CFQuart4*ReturnVol 0.017 *** 0.012 *** 0.016 *** 0.015 ** 0.018 *** 0.012 *** (0.00) (0.00) (0.01) (0.01) (0.00) (0.00) Constant 1.517 *** 1.354 *** 1.875 ** 2.108 ** 1.163 ** 0.861 (0.47) (0.42) (0.80) (0.85) (0.57) (0.51) I ndustry Dummies Y Y Y Y Y Y Y Y Y Y Y Y Number of observations 35,403 35,403 13,325 13,325 22,078 22,078 0.044 0.044 0.029 0.029 0.057 0.057 ***,**,* denote significance at the 1%, 5%, and 10% levels, respectively

PAGE 127

127 Table A 1 0 Linear combination for the choice to maintain or change repurchase level for prior repurchasers. This table shows the linear combination of coefficients for the interaction variables from Table A 9 for the firms choice to either maintain or change their repurchase level given that they repurchased stock during the prior fiscal year. The firms repurchase level is defined as purchase of common and preferred stock (Compustat Item #115) plus the minimum of 0 and (preferred stock / redemption value (Compust at Item #56) minus lag(preferr ed stock / redemption value). Standard errors are shown in parenthesis. Base choice is to maintain repurchase level. Variable Decrease Repurchase Increase Repurchase Decrease Repurchase Increase Repurchase Decrease Repurchas e Increase Repurchase Whole Sample 1962 1989 1990 2009 GARCH 0.020 0.008 0.022 0.003 0.033 *** 0.002 (0.01) (0.01) (0.02) (0.02) (0.01) (0.01) GARCH + CFQuart2*GARCH 0.020 0.002 0.007 0.023 0.027 ** 0.004 (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) GARCH + CFQuart3*GARCH 0.003 0.022 *** 0.018 0.044 *** 0.021 0.002 (0.01) (0.01) (0.01) (0.01) (0.02) (0.01) GARCH + CFQuart4*GARCH 0.022 0.033 *** 0.016 0.028 0.015 0.029 *** (0.01) (0.01) (0.01) (0.02) (0.02) (0.01) ReturnVolatility 0.008 *** 0.003 0.006 0.000 0.007 ** 0.001 (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) ReturnVolatility+ CFQuart2*ReturnVol 0.007 *** 0.001 0.010 *** 0.005 0.010 *** 0.001 (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) ReturnVolatility + CFQuart3*ReturnVol 0.016 *** 0.008 *** 0.019 *** 0.016 *** 0.010 *** 0.001 (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) ReturnVolatility + CFQuart4*ReturnVol 0.025 *** 0.015 *** 0.022 *** 0.015 *** 0.024 *** 0.013 *** (0.00) (0.00) (0.00) (0.01) (0.00) (0.00) ***,**,* denote significance at the 1%, 5%, and 10% levels, respectively

PAGE 128

128 Table A 1 1 Logit regression for the choice to maintain or increase repurchase level for prior nonrepurchasers. This table shows the logit regress ion and linear combination for interaction coefficients for the firms choice to either maintain or change their repurchase level given that they did not repurchase stock during the prior fiscal year. The firms repurchase level is defined as purchase of common and preferred stock (Compustat Item #115) plus the minimum of 0 and (preferred stock / redemption value (Compustat Item #56) minus lag(preferred stock / redemption value). Standard errors are shown in parenthesis. Base choice is to maintain not repurchasing shares. Variable Whole Sample 1962 1989 1990 2009 NYP 0.398 ** 0.094 0.517 ** (0.20) (0.16) (0.24) MtoB 0.147 *** 0.346 *** 0.146 *** (0.02) (0.06) (0.02) dA/A 0.514 *** 0.446 *** 0.547 *** (0.04) (0.06) (0.05) E/A 0.261 1.767 *** 0.082 (0.18) (0.55) (0.18) GARCH 0.028 *** 0.033 *** 0.015 (0.01) (0.01) (0.01) ReturnVolatility 0.004 *** 0.005 *** 0.003 *** (0.00) (0.00) (0.00) CashFlow/Assets 1.436 *** 0.159 1.776 *** (0.26) (0.42) (0.28) LagNetDebt/Assets 1.043 *** 0.791 *** 1.146 *** (0.05) (0.06) (0.06) NegativeRetainedEarn 0.496 *** 0.625 *** 0.489 *** (0.03) (0.09) (0.03) ChangeDivTaxRate 0.158 0.207 ** 0.317 (0.11) (0.09) (0.50) RepatTaxCutDummy 0.331 *** 0.239 ** (0.09) (0.10) PreSafeHarbor 0.441 *** 0.698 *** (0.08) (0.14) LagReturn 0.015 *** 0.027 *** 0.019 *** (0.00) (0.01) 0.006 ***,**,* denote significance at the 1%, 5%, and 10% levels, respectively

PAGE 129

129 Table A 1 1 Continued Variable Whole Sample 1962 1989 1990 2009 CFQuart2*GARCH 0.015 *** 0.025 *** 0.023 *** (0.00) (0.00) (0.01) CFQuart3*GARCH 0.026 *** 0.018 *** 0.037 *** (0.00) (0.01) (0.01) CFQuart4*GARCH 0.037 *** 0.034 *** 0.049 *** (0.01) (0.01) (0.01) CFQuart2*ReturnVol 0.003 *** 0.002 0.004 *** (0.00) (0.00) (0.00) CFQuart3*ReturnVol 0.005 *** 0.000 0.008 *** (0.00) (0.00) (0.00) CFQuart4*ReturnVol 0.007 *** 0.003 ** 0.008 *** (0.00) (0.00) (0.00) Constant 1.464 *** 0.433 ** 1.147 *** (0.27) (0.21) (0.33) Industry Dummies Y Y Y Year Clustered S.E. Y Y Y Number of observations 66,654 56,445 43,792 Adjusted R2 0.086 0.067 0.105 ***,**,* denote significance at the 1%, 5%, and 10% levels, respectively

PAGE 130

130 Table A 12 Linear combination for the choice to maintain or increase repurchase level for prior nonrepurchasers. This table shows the linear combination of coefficients for the interaction vari ables from Table A 11 for the firms choice to either maintain or change their repurchase level given that they did not repurchase stock during the prior fiscal year. The firms repurchase level is defined as purchase of common and preferred stock (Compustat Item #115) plus t he minimum of 0 and (preferred stock / redemption value (Compustat Item #56) minus lag(preferred stock / redemption value). Standard errors are shown in parenthesis. Base choice is to maintain not repurchasing shares. Variable Whole Sample 1962 1989 1990 2009 GARCH 0.028 *** 0.033 *** 0.015 (0.01) (0.01) (0.01) GARCH + CFQuart2*GARCH 0.043 *** 0.045 *** 0.039 *** (0.01) (0.01) (0.01) GARCH + CFQuart3*GARCH 0.054 *** 0.044 *** 0.053 *** (0.01) (0.01) (0.01) GARCH + CFQuart4*GARCH 0.065 *** 0.054 *** 0.065 *** (0.01) (0.01) (0.01) ReturnVolatility 0.004 *** 0.005 *** 0.003 *** (0.00) (0.00) (0.00) ReturnVolatility+ CFQuart2*ReturnVol 0.006 *** 0.006 *** 0.006 *** (0.00) (0.00) (0.00) ReturnVolatility + CFQuart3*ReturnVol 0.008 *** 0.005 *** 0.010 *** (0.00) (0.00) (0.00) ReturnVolatility + CFQuart4*ReturnVol 0.010 *** 0.007 *** 0.011 *** (0.00) (0.00) (0.00) ***,**,* denote significance at the 1%, 5%, and 10% levels, respectively

PAGE 131

131 LIST OF REFERENCES Arak, Marcelle, and Naranchimeg Mijid, 2006, The VI X and VXN volatility measures: F ear gauges or forecasts? Journal of Derivatives and Hedge Funds 12, 1427. Bagnoli, Mark, Michael Clement, and Susan Watts, 2006, Aroundthe clock m edia coverage and the t iming of e arnings a nnouncements, Working Paper Purdue University and the University of Texas at Austin. Baker, Malcolm, and Jeffrey Wurgler, 2004, A catering t heory of d ividends, Journal of Fina nce 59, 11251165. Banyi, Monica L., Edward A. Dyl, and Kathleen M. Kahle, 2008, Errors in estimating share repurchases, Journal of Corporate Finance 14, 460474. Barclay, Michael, and Terrence Hendershott, 2003, Price d iscovery and t rading a fter h ours, Review of Financial Studies 16, 1041 1073. Barclay, Michael, and Terrence Hendershott, 2004, Liquidity e xternalities and adverse selection: E vidence from t rading a fter h ours, Journal of Finance 59, 681709. Barclay, Michael J., and Jerold B. Warner, 1993, Stealth trading and volatility, Journal of Financial Economics 34, 281305. Black, Fischer, 197 6 The dividend puzzle, The Journal of Portfolio Management 2, 5 8. Blouin, Jennifer, and Linda Krull, 2009, Bringing it h ome: A study of the i ncentives surrounding the r epatriation of f oreign earnings under the A merican Jobs Creation Act of 2004, Journal of Accounting Research 47, 10271059. Brav, Alon, John R. Graham, Campbell R. Harvey, and Roni Michaely, 2005, Payout policy in the 21st C entury, Journal of Financial Economics 77, 483527. Brav, Alon, John R. Graham, Campbell R. Harvey, and Roni Michaely, 2008, Managerial r esponse to the May 2003 d ividend t ax cut, Financial Management 37, 611624. Chay, J.B., and Jungwon Suh, 2009, Payout policy and cashf low uncertainty, Journal of Financial Economics 93, 88107. Chetty, Raj, and Emmanuel Saez, 2005, Dividend t axes and c orporate b ehavior: Evidence from the 2003 dividend t ax cut, The Quarterly Journal of Economics 120, 791833. Chordia, Tarun, Richard Roll and Avanidhar Subrahmanyam, 2011 Recent trends in t rading a ctivity and market quality Journal of Financial Economics 101, 243263.

PAGE 132

132 Chung, Kee H., and Youngsoo Kim, 2009, Volatility, market structure, and the bidask spread, Asia Pacific Journal of Financial Studies 38, 67 107. Das, Somnath, Carolyn B. Levine, and K. Sivaramakrishnan, 1998, Earnings p redictability and bias in analysts e arnings f orecasts, The Accounting Review 73, 277294. DeAngelo, Harry, Linda DeAngelo, and Douglas J. Skinner, 2004, Are dividends disappearing? Dividend concentration and the consolidation of earnings, Journal of Financial Economics 72, 425456. Engle, Robert, 2001, GARCH 101: The use of ARCH/GARCH models in applied econometrics, Journal of Economic Perspectives, 15, 157 168. Fama, Eugene F., and Kenneth R. French, 2001, Disappearing dividends: changing firm characteristics or lower propensity to pay? Journal of Financial Economics 60, 343. Fuller, Kathleen P., and Michael A. Goldstein, 20 11, Do d ividends m atter more in d eclining m arkets? Journal of Corporate Finance 17, 457473 Greene, Jason, and Susan Watts. 1996, Price d iscovery on the NYSE and the NASDAQ : The case of o vernight and d aytime n ews r eleases, Financial Management 25, 1942. Grullon, Gustavo, and Roni Mi chaely, 2002, Dividends, s hare r epurchases, and the substitution h ypothesis, Journal of Finance 57, 1649 1684. Hao, Jinji, and Jin E. Zhang, 2010, GARCH option pricing models, the CBOE VIX and variance risk premium, Working Paper, Peking University and the University of Hong Kong. Hoberg, Gerard, and Nagpurnanand R. Prabhala 2009, Disappearing dividends, catering and r isk, Review of Financial Studies 22, 79116. Houston, Joel F., and Michael D. Ryngaert, 1992, The links between trading time and market volatility, Journal of Financial Research 15, 91100. Jain, Prem C., and Gun Ho Joh, 1988, The dependence between h ourly p rices and t rading v olume, Journal of Financial and Quantitative Analysis 23, 269283. Lim, Terence, 2001, Rationality and analysts f orecast b ias, Journal of Finance 56, 369385. Lintner, John, 1956, Distribution of incomes of c orporations among d ividends, r etained e arnings, and t axes, The American Economic Review 46, 97113.

PAGE 133

133 Miller, Merton, and Franco Modigliani, 1961, Dividend p olicy, growth and the v aluation of shares, Journal of Business 34, 411433. Neumark, David, P.A. Tinsley, and Suzanne Tosini, 1991, After h ours stock prices and p ost crash h angovers, Journal of Finance 46, 159178. New York Stock Exchange, Inc, 2008, TAQ Users Guide, Version 1.1.9, Retrieved April 26, 2011, from http://wrds web.wharton.upenn.edu/wrds/support/Data/_001Manuals%20and%20Overviews/_ 125TAQ/TAQ%20User%27s%20Guide%2010 2008%20edition.pdf.cfm Oldfield, George S., and Richard J. Rogalski, 1980, A theory of common stock returns over trading and nontrading periods, Journal of Finance 35, 729751. Skinner, Douglas J., 2008, The evolving relation between earnings, dividends, and stock repurchases, Journal of Financial Economics 87, 582609. Stoll, Hans R., 1989, Inferring the c omponents of the b id a sk s pread: t heory and e mpirical t ests, Journal of Fi nance 44, 115134. Van Bommel, Jos, 2009, Measuring price discovery: The variance ratio, the R2 and the Weighted Price Contribution, Working Paper, Universidad Cardenal Herrera. Wang, Jianxin, and Minxian Yang, 2010, How well does the Weighted Price Contri bution measure price discovery? University of Technology Sydney and University of New South Wales, Working Paper. Whaley, Robert E., 2000, The i nvestor f ear g auge, The Journal of Portfolio Management 26, 1217. Zdorovtsov, Vladimir, 2003, Firm Specific News, Extendedhours t rading and v ariances over t rading and n ontrading p eriods, Working Paper University of South Carolina.

PAGE 134

134 BIOGRAPHICAL SKETCH Brian received his Bachelor of Science in m athematics with a minor in f inance from Trevecca Nazarene University in Nashville, T ennessee. He then earned his Master of Business Administration in f inance and m anagement from the Crummer Graduate School of Business at Rollins College in Winter Park F lorida Brian also worked in the research department of CNL Income Corp. in Orlando while attending Crummer. After finishing his MBA degree, Brian began Ph.D. studies in Finance at the University of Florida. After completing his Ph.D in 2011 Brian accepted a position as Assistant Professor of Finance at the University of Tulsa in Tulsa, O klahoma